US12175976B2 - Multi-assistant device control - Google Patents

Abstract

A speech-processing system may provide access to one or more virtual assistants via an audio-controlled device. A multi-agent component of a device may be used to coordinate control of device process(es) regardless of the invoked agent and without necessarily sharing information between agents and corresponding speech-processing system(s). Device-state information may be sent from the multi-agent component to a speech-processing system to interpret a user command. If the command relates to a device process control it may be routed to a device skill and to the multi-agent component of the device. The multi-agent component may track a device state and may inform a particular speech-processing system if a process related to the particular speech-processing system has been controlled.

Description

CROSS-REFERENCE TO RELATED APPLICATION

This application claims priority to U.S. Provisional Patent Application No. 63/400,636, filed Aug. 24, 2022 and titled “MULTI-ASSISTANT DEVICE CONTROL,” the content of which is expressly incorporated herein by reference in its entirety.

BACKGROUND

Speech recognition systems have progressed to the point where humans can interact with computing devices using their voices. Such systems employ techniques to identify the words spoken by a human user based on the various qualities of a received audio input. Speech recognition combined with natural language understanding processing techniques enable speech-based user control of a computing device to perform tasks based on the user's spoken commands. Speech recognition and natural language understanding processing techniques may be referred to collectively or separately herein as speech processing. Speech processing may also involve converting a user's speech into text data which may then be provided to various text-based software applications.

Speech processing may be used by computers, hand-held devices, telephone computer systems, kiosks, and a wide variety of other devices to improve human-computer interactions.

BRIEF DESCRIPTION OF DRAWINGS

For a more complete understanding of the present disclosure, reference is now made to the following description taken in conjunction with the accompanying drawings.

FIG.1 is a conceptual diagram illustrating components of a virtual assistant system with features for protected cross-assistant command processing, according to embodiments of the present disclosure;

FIGS.2A-2B are signal flow diagrams illustrating example operations for protected cross-assistant command processing, according to embodiments of the present disclosure;

FIGS.3A-3B are flowcharts illustrating example operations for performing cross-assistant command processing, according to embodiments of the present disclosure;

FIG.4 is a conceptual diagram of components of a speech processing system, according to embodiments of the present disclosure;

FIG.5 is a conceptual diagram illustrating components that may be included in a device, according to embodiments of the present disclosure;

FIG.6 is a conceptual diagram of an automatic speech processing component, according to embodiments of the present disclosure;

FIG.7 is a conceptual diagram of how natural language processing is performed, according to embodiments of the present disclosure;

FIG.8 is a conceptual diagram of how natural language processing is performed, according to embodiments of the present disclosure;

FIG.9 is a conceptual diagram of text-to-speech components according to embodiments of the present disclosure;

FIG.10 is a block diagram conceptually illustrating example components of a device, according to embodiments of the present disclosure;

FIG.11 is a block diagram conceptually illustrating example components of a system, according to embodiments of the present disclosure; and

FIG.12 illustrates an example of a computer network for use with the overall system, according to embodiments of the present disclosure.

DETAILED DESCRIPTION

Speech processing systems and speech generation systems can be combined with other services to create virtual “assistants” that a user can interact with using natural language inputs such as speech, text inputs, or the like. The assistant can leverage different computerized voice-enabled technologies. Automatic speech recognition (ASR) is a field of computer science, artificial intelligence, and linguistics concerned with transforming audio data associated with speech into text or other type of word representative data of that speech. Similarly, natural language understanding (NLU) is a field of computer science, artificial intelligence, and linguistics concerned with enabling computers to derive meaning from text or other natural language meaning representation data. ASR and NLU may be used together as part of a speech processing system, sometimes referred to as a spoken language understanding (SLU) system. Text-to-speech (TTS) is a field of computer science concerning transforming textual and/or other meaning representation data into audio data that is synthesized to resemble human speech. ASR, NLU, and TTS may be used together to act as a virtual assistant that can respond to spoken commands and respond with synthesized speech. For example, an audio-controlled user device and/or one or more speech-processing systems may be configured to receive human speech and detect a wakeword used to activate the device and/or other natural language input. The device and/or system may determine a command represented by the user input, and use TTS and/or other system command to provide a response (e.g., in the form of synthesized speech, command to send audio to a different device/system component, etc.).

Some audio-controlled devices can provide access to more than one speech-processing system, where each speech-processing system may provide services associated with a different virtual assistant. In such multi-assistant systems, one or more of the speech-processing systems may be associated with its own set of wakewords, which for invoking the speech-processing system, as well as associated with other observable characteristics such as voice characteristics and other audible or visual indicators that allow a user to identify which speech-processing system the user is interacting with.

Under certain circumstances an overall system may be configured in a manner that does not allow certain communications/operations between speech-processing systems. This may be for a number of reasons. First, a user may have one set of permissions with a first speech-processing system that does not allow sharing of certain data to a second speech-processing system. Further, centralized user settings may not permit sharing of user utterance information to a speech-processing system that was not directly invoked by the user. Thus, to increase the perceived protection of user privacy, communication between different speech-processing system components may be prevented. Second, while a first speech-processing system and a second speech-processing system may each be invoked from a first device, the systems themselves may not desire to share information directly between them. This may true be in a situation where the first speech-processing system and second speech-processing system have different speech processing architectures/pipelines (as opposed to sharing many speech processing components) such that commands invoking the first speech-processing system go to one set of device(s)/component(s) for processing while commands invoking the second speech-processing system go to a different set of device(s)/component(s) for processing. This may be true in the situation where a first speech-processing system and second speech-processing system are managed by competing entities. Thus it may not be permitted to share information directly between different speech-processing system components. Thus, in certain overall system configurations components of a first speech-processing system may not be configured to communicate with components of a second speech-processing system.

In certain circumstances, however, a user may speak a command using the wakeword of one system when the command actually relates to a different system. For example, a user may start a timer using a wakeword of a first assistant/system (for example, “Alexa, set a timer for 10 minutes”) but, when the timer has ended and a device is beeping, the user may try to stop the timer using the wakeword of a second assistant/system (for example, “Kitchen, cancel the timer”). If a first speech-processing system (such as one associated with the wakeword “Alexa”) is separate from a second speech-processing system (such as one associated with the wakeword “Kitchen”), the user's command to stop the timer may not be understood by the second speech-processing system as it would not necessarily process the command “cancel the timer” correctly as it would have no information about any timer in progress since the initial command to set the timer was processed by the first speech-processing system. In such a circumstance the second speech-processing system may return an error to the user (for example, output audio of “I'm sorry, there are no timers set”) which may cause user frustration, particularly if the user may not recall which wakeword the user spoke when starting the timer and thus may be unsure how to stop the beeping.

In another example, a first user may start a process (e.g., a timer, playing media, turning on a camera feed, etc.) using a first assistant/system but a second user may attempt to control the process (e.g., by stopping the process, skipping a song, etc.) using a second assistant/system. Such a situation may occur with a shared device where the first user interacts with the shared device using the first assistant while the second user interacts with the shared device using the second assistant. If the first user starts a process with the device and leaves the room, the second user may experience difficulty controlling the ongoing process using the second assistant. This may occur because, similar to the example above with a timer, the second assistant may not have information regarding the process which may result in output of an error and an undesired user experience.

Offered are techniques and components to improve the ability to process commands across speech-processing systems, particularly when those speech-processing systems may have limits that govern what information may be exchanged directly between them. A device may include a multi-assistant component that can pass information and commands and/or arbitrate the use of device resources between components that are dedicated to a particular speech-processing system. Individual speech-processing system(s) may also have information indicating which commands relate to device control (for example which commands should be handled by a device-control skill) so that commands related to device control may be processed by one system even if they may relate to device controls of another system. Although the disclosure below illustrates operations with regard to two speech-processing systems, the teachings herein are applicable to configurations with more than two speech-processing systems.

Speech-processing systems may be configured to determine when a user input requests control of a device process. In such circumstances, the system will send data for the request to a special device skill that can communicate with a dedicated component on the device that can manage device process commands, communicate with multiple speech-processing systems, and arbitrate between them. The device skill will send the command to the dedicated component on the device, thus enabling control of a device process by one speech-processing system, even if the device process may have been initiated by a different speech-processing system.

A device process may involve controlling a process that involves some action to be performed by a device. Such a device process control may include, for example, starting/stopping a timer, setting/stopping an alarm, playing/stopping media content (such as a song, video, podcast, etc.), controlling output content (such as skipping a song, going back a song, extending/snoozing a timer/alarm, stopping synthesized speech output, etc.), setting a temperature (for example if a device may operate as a thermostat), activating/deactivating a component of the device (such as a camera, light, etc.), controlling a device setting (such as volume, brightness, sensitivity, etc.), setting/controlling a reminder, initiating/controlling/terminating a call or call request, or the like. A device process control may thus control a device to transition from a first state (e.g., outputting audio, showing something on a display) to a second state (e.g., ceasing output of audio, outputting audio at a different volume, showing something else on the display, removing something from the display, etc.).

Although the respective systems are referred to herein as “speech-processing systems” they may also be considered natural language processing systems in that they may be configured to process natural language inputs that may not necessarily be spoken and may be input using some other method such as text inputs to an application (or the like) where the application may correspond to a particular assistant/system. Thus, inputs and outputs from the device need not be in (or represent) spoken language. In some implementations, the user may be able to input natural language inputs via text, braille, American Sign Language (ASL), etc., depending on system configuration. Other inputs to trigger a processing system are also possible, such as an acoustic event (e.g., baby crying, footsteps), a button press, etc.

The system may be configured to incorporate user permissions and may only perform activities disclosed herein if approved by a user. As such, the systems, devices, components, and techniques described herein would be typically configured to restrict processing where appropriate and only process user information in a manner that ensures compliance with all appropriate laws, regulations, standards, and the like. The system and techniques can be implemented on a geographic basis to ensure compliance with laws in various jurisdictions and entities in which the components of the system and/or user are located.

FIG.1 is a conceptual diagram illustrating components of avirtual assistant system100 with features for protected cross-assistant command processing, according to embodiments of the present disclosure. Thevirtual assistant system100 may include an audio-enableddevice110, a first natural language/speech-processing system120a(which may be abbreviated “first system120a”), and a second natural language/speech-processing system120b(which may be abbreviated “second system120b”). Thefirst system120aand thesecond system120bmay be referred to collectively as “systems120.” AlthoughFIG.1 illustrates thefirst system120aand thesecond system120bas having similar components in a similar arrangement, the components, functions, and/or architectures of thefirst system120aand thesecond system120bmay differ. In addition, some or all of the components and/or functions of one or both of thefirst system120aand/or thesecond system120bmay reside on, or be performed by, thedevice110. Other possible arrangements of the components and functions of thedevice110 and thesystems120 are described in additional detail below with reference toFIGS.4 and5. As noted above, while twosystems120 are shown for sake of illustration, any number and variety ofsystems120 can be supported and be coordinated consistent with the principles and processes described herein.

Thedevice110 may receive audio corresponding to a spoken natural language input originating from a user (not illustrated). Thedevice110 may process audio following detection of a wakeword. A wakeword may be a word or phrase that, when detected, may cause adevice110 to invoke a speech-processing system120 for processing audio data that accompanies or includes the wakeword. The wakeword may be specific to a particular speech-processing system120. Thus if adevice110 detects a first wakeword it may route data corresponding to the speech to a first speech-processing system while if thedevice110 detects a second wakeword it may route data corresponding to the speech to a second speech-processing system. (Thedevice110 may also be configured to detect any number of wakewords having any correlation with the set of available speech-processingsystems120 such that no wakeword is associated with more than one speech-processing system120.) Thedevice110 may generate audio data corresponding to the audio/speech, and may send the audio data to thefirst system120aand/or thesecond system120b. Thedevice110 may send the audio data to thesystems120 via one or more applications installed on thedevice110. An example of such an application is the Amazon Alexa application that may be installed on a smart phone, tablet, or the like. In some implementations, thedevice110 may receive text data corresponding to a natural language input originating from the user5, and send the text data to one of thesystems120. Thedevice110 may receive output data from thesystem120, and generate a synthesized speech output and/or perform some action. Thedevice110 may include a camera for capturing image and/or video data for processing by thesystems120. Examples ofvarious devices110 are further illustrated inFIG.12.

Thesystems120 may include supporting components of local and/or remote system(s) such as a group of computing components located geographically remote fromdevice110 but accessible via network199 (for example, servers accessible via the internet). Thesystems120 may also include a remote system that is physically separate fromdevice110 but located geographically close todevice110 and accessible via network199 (for example a home server located in a same residence as device110). Thesystems120 may also include some combination thereof, for example where certain components/operations are performed via on device components or a home server(s) and others are performed via a geographically remote server(s). Although the figures and discussion of the present disclosure illustrate certain steps in a particular order, the steps described may be performed in a different order (as well as certain steps removed or added) without departing from the present disclosure.

Thedevice110 may include amicrophone114 for receiving audio and aspeaker112 for emitting audio. Thedevice110 may include one ormore wakeword detectors121 capable of detecting one or more wakewords. In some implementations, awakeword detector121 may be embedded in a processor chip; for example, a digital signal processor (DSP). In some implementations, awakeword detector121 may be an application-driven software component. In certain instances asingle wakeword detector121 may be capable of detecting multiple wakewords for more than one system. In other instances adevice110 may include multiple wakeword detectors, such as afirst wakeword detector121aand asecond wakeword detector121b, each capable of detecting its own wakeword. For example, afirst wakeword detector121amay detect one or more wakewords associated with thefirst system120a, and asecond wakeword detector121bmay detect one or more wakewords associated with thesecond system120b.

The device may include one or more assistant components140 including thefirst assistant component140aand thesecond assistant component140b. The assistant component(s)140 may interface with one or more of thesystems120. In theexample system100 shown inFIG.1, thefirst assistant component140acommunicates with thefirst system120a, and thesecond assistant component140bcommunicates with thesecond system120b. In some implementations, a single assistant component140 may handle communications with more than onesystem120. Thedevice110 may have a dedicated assistant component140 for asystem120, or a single assistant component140 communicating with allsystems120. The device may include amulti-assistant component115 for managing multi-assistant and cross-assistant operations of thedevice110 as described herein. The device may also include a set of components to store/track state data194. (As noted below,state data194 can be separately tracked and maintained by each assistant component140 as well as by themulti-assistant component115.)Such state data194 may indicate the state of the device110 (and/or a user profile corresponding to the device110) and may correspond to one or more processes of the device. Examples of state data may include volume level, data indicating what is being shown on a display, time data, network access data, timer status, or the like. Thestate data194 may be stored on thedevice110 or potentially on another device such as a remote device, home server, or the like. Additional components of thedevice110 are described in additional detail below with reference toFIG.10.

In certain configurations, to maintain privacy perception and/or other separation between speech-processing systems, afirst assistant component140amay not be configured to communicate with asecond assistant component140bwithout routing the communication through themulti-assistant component115. In this way themulti-assistant component115 may mediate the interactions between the speech-processing system components. Similarly, the multi-assistant component115 (or other remote/cloud component) may mediate communications between thefirst system120aand thesecond system120b. Thus speech-processing systems may not be configured to directly communicate, particularly when such communications may involve a particular utterance being processed. While illustrated to operate physically ondevice110, themulti-assistant component115 may operate on a different physical device, for example a home server or the like. In such (or other) situations themulti-assistant component115 may coordinate multi-assistant operations formultiple devices110, wheresuch devices110 may be associated with one or more user accounts. For example, a singlemulti-assistant component115 may coordinate multi-assistant operations for multiple device(s) associated with a particular user/user profile, family/family profile/multiple user profile(s), or the like.

As part of such separation, in certain configurations, each speech-processing system and/or components associated therewith, may store/manage theirown state data194 with respect to the device. For example, afirst assistant component140aassociated withfirst system120amay store/managestate data194awhich includes data regarding interactions/operations with regard to the device110 (and/or a user profile associated with device110) andfirst system120a. For example, if a user interacts withdevice110 to invokefirst system120a(for example by speaking a first wakeword associated withfirst system120a), thefirst assistant component140amay save certain information regarding the interaction betweendevice110 andfirst system120aasstate data194a. Thus, if a device process is initiated as a result of a command tofirst system120a, thefirst assistant component140amay store information regarding that device process asstate data194a. For example, if a user starts a timer by invoking a first assistant associated withfirst system120a, thestate data194amay reflect the start time of the timer, time remaining, label associated with the timer, etc. In another example, if a user starts to play music by invoking a first assistant associated withfirst system120a, thestate data194amay reflect the start of the music, the source of the music content (e.g., music service), information about currently playing music, information about previously played music, etc.

Asecond assistant component140bmay also store/manage itsown state data194bwith respect to interactions/operations with regard to the device110 (and/or a user profile associated with device110) andsecond system120b. Such management ofstate data194bwith regard tosecond system120bmay operate similarly to that described above with regard to194aandfirst system120a. As part of the separation of systems, however,first assistant component140amay not have access tostate data194bandsecond assistant component140bmay not have access tostate data194a. Thus each system/assistant component may only trackstate data194 with respect to its own operations.

Certain state data194 may also be stored/managed bymulti-assistant component115. Such state data is shown inFIG.1 as194m. Thisstate data194mmay include information related to device processes ongoing at the device and may include some portion(s) of information stored in194a/194band/or other information about management of device processes. For example, if a timer is ongoing,state data194mmay include an indicator that the timer is ongoing and thesystem120 that was used to invoke the timer, but may not include as many details of the timer as thestate data194 of the invoking system. Similarly, if music is being output,state data194mmay indicate that music is being played but may not include all the details of the music playback.State data194mmay indicate which device process(es) are active at any particular point in time (for example, timer ongoing, timer ended and beeping, music playback ongoing, etc.)State data194mmay indicate which device controls are executable for a particular device process (whether ongoing or not). For example,state data194mmay indicate if adevice110 is capable of stopping, extending, pausing a timer; stopping, pausing, adjusting volume for music playback, etc.State data194mmay also indicate which channels (e.g., hardware components) are currently being used by what process(es), etc. Information may be exchanged betweenmulti-assistant component115 and a single assistant component140 to update the respective state data(s) and/or execute controls for thedevice110/a device process. For example, an application programming interface (API) or other interface, registration process, etc. may be used to coordinate between themulti-assistant component115 and a single assistant component140 to exchange information about a state/process of thedevice110.

Thesystems120 may include various components for processing natural language commands. Asystem120 may include a language processing component192 for performing operations related to understanding natural language such as ASR, NLU, entity resolution, etc. Thesystem120 may include a language output component193 for performing operations related to generating a natural language output, such as TTS. Thesystem120 may also include a component to track system state data195. Such system state data195 may indicate the state of operations of therespective system120 for example with respect to aparticular device110, user profile, or the like. For example, state data195 may include dialog data, indications of previous utterance(s), whether thesystem120 has any ongoing processes for thedevice110/user profile, or the like. Thesystem120 may include one or more skill components190. The skill components190 may perform various operations related to executing commands such as online shopping, streaming media, controlling smart-home appliances, and the like.

One of the skills available to thesystem120 may include a device skill191. Such a device skill may be configured to handle and manage specific utterances that involve controlling a device process or a device state. Eachsystem120 may have its own device skill191 and/or a central device skill191 may be accessible tomultiple systems120. Each device skill191 may be associated with its own skill processing components125 (discussed below).

A device skill191 may be configured to communicate with thedevice110 through the assistant component(s)140. Thus the device skill191 may send commands to control the device110 (in coordination with the multi-assistant component115) through the assistant component(s)140. Thus asystem120 may send a command/message/directive to themulti-assistant component115 by routing such communication(s) through individual assistant components140. For example, a device control command fromfirst system120amay be sent fromdevice skill191ato thefirst assistant component140awhich then routes the device control command to themulti-assistant component115.

FIGS.2A-B are signal flow diagrams illustrating example operations for protected cross-assistant command processing, according to embodiments of the present disclosure.FIG.2A illustrates operations in which a user can instruct afirst system120ato initiate a device process.FIG.2B illustrates operations in which a user can instruct asecond system120bto control the same device process that was initiated inFIG.2A using thefirst system120ausing components such as115,191, etc. rather than by allowing communications directly between thefirst system120aand thesecond system120b. Specifically,FIG.2A illustrates operations between amicrophone114, aspeaker112, awakeword detector121, afirst assistant component140a, amulti-assistant component115, and asecond assistant component140bof adevice110, and thefirst system120a,device skill191a(which may be of thefirst system120a), and thesecond system120b.

As shown inFIG.2A, themicrophone114 may receive an audio signal and send (202) audio data to thewakeword detector121. The audio data may represent, for example, a natural language command such as: “Alexa, set a timer for 10 minutes.” Thewakeword detector121 may detect the wakeword “Alexa,” corresponding with the first speech-processing system120aand thefirst assistant component140a. Thewakeword detector121 may notify (204) themulti-assistant component115 that the first wakeword was detected in the input.

In some implementations, thedevice110 may receive input data in other formats, such as typed or scanned text, braille, or American Sign Language (ASL) (for example as detected by processing image data and/or sensor data representing a user communicating in ASL). Thedevice110 may determine that the input data is to be processed by thefirst system120abased on other indications, such as a button press or because thefirst system120arepresents adefault system120 for executing commands from thedevice110.

Themulti-assistant component115 may signal (206) thefirst assistant component140athat thefirst assistant component140amay send data representing the command (e.g., the audio data) to thefirst system120a. After themulti-assistant component115 confirms invocation of thefirst system120a(e.g., through detection of the first wakeword by the wakeword detector121), the audio data of the utterance may be sent to thefirst assistant component140a. The audio data may be sent (207) to thefirst assistant component140aby themulti-assistant component115 or by another component. Themulti-assistant component115 may also send (208) thefirst assistant component140astate data194 corresponding to the state of thedevice110 and/or the user profile corresponding to thedevice110. The state data being sent (208) may be data available to the multi-assistant component115 (e.g., taken fromstate data194m). Such state data may indicate one or more process controls capable of being executed bydevice110. In certain instances, such state data being sent may only apply to active process(es) of thedevice110. Thus the state data sent may comprises metadata corresponding to process(es) and/or process controls of thedevice110. Thefirst assistant component140amay send (210) audio data representing the command to thefirst system120a. Thefirst assistant component140amay also send (212) the state data to thefirst system120a. The state data sent (212) from thefirst assistant component140ato thefirst system120amay include all or some of the state data (e.g., portion ofstate data194m) sent from themulti-assistant component115 to thefirst assistant component140aabove instep208. The state data sent (212) from thefirst assistant component140ato thefirst system120amay also include certain state data (e.g.,state data194a) available to thefirst assistant component140athat may be distinct from thestate data194m. The state data may indicate one or more active hardware component(s) of thedevice110, which may indicate activity on certain processing channel(s) of the device. The system may use a particular identifier, such as an utterance identifier to track data related to the utterance across the different components. For example, as audio data of an utterance is received thedevice110 may assign a particular identifier to the audio data. That identifier may be sent along with the WW detection indication (204), the WW signal (206), state data (208/212), audio data (210), etc. so that thedevice110,system120a, etc. can keep track of information related to the particular utterance.

The device/profile state data194 (either corresponding to one of the assistant component(s)140 and/or the multi-assistant component115) ultimately sent to a speech-processing system120 may be obtained from storage on thedevice110 or from storage on a different device corresponding to a same user profile. The state data194 (in particular thestate data194m) may indicate what processes are ongoing and/or controllable using the device110 (and/or other device associated with a user profile associated with the state data194). For example, if thedevice110 has ongoing dialog and timer thestate data194mmay indicate <dialog>; <timer>. Thestate data194mmay also indicate what commands may be performed (which may relate to a device process). For example, thestate data194mmay indicate <stopdialog>; <stoptimer>; <adjusttimer> or the like. Thestate data194mmay indicate a number of different controllable device process(es) or may only indicate active device process(es). Thestate data194mmay also indicate a type of the command. For example, certain commands may relate to dialog control, alert control, content control, etc. Such types may be indicated in thestate data194m. Thestate data194mmay also indicate priority information relative to particular device process(es). For example, if a device is playing music while also outputting a beeping corresponding to a timer that has expired, thestate data194mmay indicate the timer has a higher priority than the music playback. Depending on system configuration, user permissions/privacy settings, or the like, the state data sent (208/212) may be limited in some manner, for example only indicating ongoing device process(es), to reduce the amount of state data shared.

Thefirst system120amay perform speech processing (for example usinglanguage processing components192aand corresponding operations described herein) to determine if the input user request corresponds to a request to control a device process and thus should be routed (214) to a device skill. Such a determination may involve processing both the audio data and other data indicating which potential commands/requests may relate to control of a device process. That other data may include the state data (e.g., state data194) that corresponds to a device/user profile. Thestate data194 may indicate what commands correspond to a device process, thus allowing thefirst system120ato properly determine when an incoming request corresponds to a device process, thus indicating whether processing for that request should be handled by a device skill. Such a device skill may be configured to send commands to the multi-assistant component115 (using routing discussed herein) for purposes of indicating to the device one or more directive(s) to control device process(es). Such directives allow themulti-assistant component115 to coordinate control of the device with assistant component(s)140 and other components. AlthoughFIGS.2A (and2B) illustrate the device/profile state data194 being sent from themulti-assistant component115, thefirst system120amay also obtain device/profile state data194 from another source, for example an off-device storage component that is in communication with thefirst system120amay provide access to the state data, which may be similar tostate data194m(or other state data194). Such a determination may also involve processing of other data indicating the process capabilities of thedevice110 and/or other device(s) corresponding to the user profile.

If the incoming request does correspond to a device skill (e.g., the request corresponds to a control of a device process) as determined by the language processing (for example using other data/state data194), thefirst system120amay route (216) data related to the request to thedevice skill191aof the first system. Such data may include NLU results data (such asNLU output data885/rankedoutput data825 as discussed below) and/or may include processed data that is based on such NLU results data but has been converted into a particular command specific for the device process.

In the example of the utterance of “Alexa, set a timer for 10 minutes,” thefirst system120amay usestate data194 to determine that adevice110 is capable of operating a timer and may determine that the command corresponds to a request to control a device process and may determine that the device skill should be invoked. Thus thefirst system120amay send to thedevice skill191adata representing the request, for example an instruction to start a timer, or the like. Thedevice skill191amay then generate (218) output data to be included in a message to the device. Thedevice skill191amay send (220) the message data back to thedevice110. Such a message may be routed through thefirst assistant component140a. The message data may include a directive to control a device process, the example ofFIG.2A that may be a directive to start a timer. Thefirst assistant component140amay receive the message data and may determine that it includes a directive corresponding to a device process control. The message data may also (or in the alternative) include information indicating that the message corresponds to a process to be managed by themulti-assistant component115. Thefirst assistant component140amay process the message data to determine (based on the directive and/or routing instructions) that the request should be handled by themulti-assistant component115. Thefirst assistant component140amay then route the message/directive data (or a portion thereof) (223) to themulti-assistant component115. Thus, as the message data may be coming from a component (thedevice skill191a) associated with thefirst system120ait may be first routed through thefirst assistant component140abefore being sent to themulti-assistant component115.

The message/output data may include the utterance identifier that links the particular output data to the input data (e.g., the audio data202). In this manner, thefirst system120amay indicate to thedevice110 that the particular message/output data goes with a particular input that thedevice110 received. The message/output data may include a directive/command to thedevice110 to begin a 10 minute timer. The directive may also include an indication as to the assistant/system that the user invoked during the request (e.g., “Alexa”) so that the appropriate system may keep track of the timer for management purposes. Thus, in the example ofFIG.2A, the message data may include an indication of thefirst system120a.

The message data may also include data to be output as an acknowledgement of the command, such as synthesized speech acknowledging the command, data to be shown on a display of thedevice110 acknowledging the command, or the like. If such acknowledgement is included, it may be output by the device, for example by sending (222) output audio of the acknowledgement message from thefirst assistant component140a(or other component) to thespeaker112 for output as synthesized speech (e.g., “starting your timer now”).

Thefirst assistant component140aand/ormulti-assistant component115 may then execute (224) the directive that was received from thedevice skill191a. In the above example, this may include executing the request to start a timer. Themulti-assistant component115 may receive the directive/message data and process it to determine the requested control is one executable by thedevice110. Themulti-assistant component115 may also determine that the process to be controlled relates tofirst assistant component140a. For example, the request may also indicate the assistant that the user originally invoked for the timer. Alternatively (or in addition) the request may indicate the identifier for the original utterance. That identifier may be used by themulti-assistant component115 to identify the wakeword that accompanied the original utterance, thus allowing themulti-assistant component115 to determine the originally invoked assistant. Themulti-assistant component115 may then instruct thefirst assistant component140a(and/or other component(s)) to take actions necessary to start the timer (or otherwise execute the directive to control the device process). Thefirst assistant component140amay then take actions to execute the directive (e.g., start the timer) and may indicate to themulti-assistant component115 that the timer has started. For example thefirst assistant component140amay register with themulti-assistant component115 information about the timer control, for example what controls are available with respect to the timer, and/or other timer data.

Themulti-assistant component115 may then update its device/userprofile state data194mto note data related to the timer. For example, themulti-assistant component115 may update (225)state data194mto indicate that a timer is active and that thedevice110 may be configured to respond to commands controlling/stopping the particular active timer. Thefirst assistant component140amay also update (226) itsown state data194ato indicate information about the timer and may send (228) a message to thefirst system120aregarding starting of the requested timer. Thefirst system120amay then take any actions needed related to the starting of the timer and may update (230) its own state data locally and remotely (e.g.,system state data195a) to indicate the requested timer (e.g., length of timer, timer start time, timer end time, associated device and/or user profile), etc.

In this manner, as reflected inFIG.2A, a user may invoke a first assistant to initiate a command that controls a device process (e.g., the setting of a 10 minute timer). The invokedfirst system120amay process the request, determine it relates to controlling a process of thedevice110, and route information about that request through themulti-assistant component115, which in turn may manage the process with thefirst system120a.

As can be seen inFIG.2B, that device process may also be controlled using a second invoked assistant that is different from the original invoked assistant, even if the respective systems for those assistants are not in direct communication or aware of the other assistant being accessible via thedevice110. Themulti-assistant component115 may allow such operations as described herein.

Continuing the example illustrated inFIG.2A, after 10 minutes the timer may expire and thedevice110 may be outputting audio corresponding to the end of the timer such as a beeping, or the like. In the example ofFIG.2B a user (which may or may not be the same user that initiated the timer) may speak a command to end the timer, only accompanied by a different wakeword (for a different assistant/system) than the original request to start the timer. For example, the command to terminate the timer may be “Kitchen, turn off the timer” where “Kitchen” is a second wakeword associated with a second assistant/second system120bthat is different from the first assistant/first system120a.

As shown inFIG.2B, themicrophone114 detects the audio of the utterance and sends (232) audio data to thewakeword detector121. Thewakeword detector121 may detect the wakeword “Kitchen,” corresponding with the second speech-processing system120band thesecond assistant component140b. Thewakeword detector121 may notify (234) themulti-assistant component115 that the second wakeword was detected in the input.

Themulti-assistant component115 may signal (236) thesecond assistant component140bthat thesecond assistant component140bmay send data representing the command to thesecond system120b. After themulti-assistant component115 confirms invocation of thesecond system120b(e.g., through detection of the second wakeword by the wakeword detector121), the audio data of the utterance may be sent to thesecond assistant component140b. The audio data may be sent (237) to thesecond assistant component140bby themulti-assistant component115 or by another component. Themulti-assistant component115 may also send (238) thesecond assistant component140bstate data194 corresponding to the state of thedevice110, including processes and state data shared by other assistants, and/or the user profile corresponding to thedevice110. The state data being sent (238) may be data available to the multi-assistant component115 (e.g., taken fromstate data194m). Such state data may indicate one or more process controls capable of being executed bydevice110. In certain instances, such state data being sent may only apply to active process(es) of thedevice110. Thus the state data sent may comprises metadata corresponding to process(es) and/or process controls of thedevice110. In the example ofFIG.2B, the state data being sent may indicate the active timer and one or more controls executable by thedevice110 with respect to the timer (e.g., stop timer, pause timer, etc.) Thesecond assistant component140bmay send (240) audio data representing the command to thesecond system120b. Thesecond assistant component140bmay also send (242) the state data to thesecond system120b. The state data sent (242) from thesecond assistant component140bto thesecond system120bmay include all or some of the state data (e.g., portion ofstate data194m) sent from themulti-assistant component115 to thesecond assistant component140babove instep238. The state data sent (242) from thesecond assistant component140bto thesecond system120bmay also include certain state data (e.g.,state data194b) available to thesecond assistant component140bthat may be distinct from thestate data194m. The state data may indicate one or more active hardware component(s) of thedevice110, which may indicate activity on certain processing channel(s) of the device. As with the first utterance ofFIG.2A, the system may use a particular identifier, such as an utterance identifier to track data related to the second utterance across the different components. That second identifier may be sent along with the WW detection indication (234), the WW signal (236), state data (238/242), audio data (240), etc. so that thedevice110,system120b, etc. can keep track of information related to the second utterance.

Thestate data194 sent (238) to thesecond assistant component140band/or sent (242) to thesecond system120bmay indicate available timer controls, that the originally requested timer has expired and is causing output of corresponding audio (e.g., the beeping), or similar state information. In some implementations, the state data that is shared by themulti-assistant component115 can be selected to be as minimal as necessary for each assistant to be aware of the shared state data from other assistants, such as a listing of shared controls, processes, focus channels, and the like. As noted above with regard to the example ofFIG.2A, in the example ofFIG.2B, thesecond system120bmay obtain state data from the device110 (e.g., through thesecond assistant component140b) and/or from another source ofstate data194.

Thesecond system120bmay perform speech processing (for example usinglanguage processing components192band corresponding operations described herein) to determine if the input user request corresponds to a request to control a device process and thus should be routed (244) to a device skill. Such a determination may involve processing both the audio data and other data indicating which potential commands/requests may relate to control of a device process. That other data may include the state data (e.g., state data194) that corresponds to a device/user profile including that state data that was shared by other assistants via the multi-assistant component. Such a determination may also involve processing of other data indicating the process capabilities of thedevice110 and/or other device(s) corresponding to the user profile.

In the example ofFIG.2B, thesecond system120b/language processing determines the incoming request does correspond to a device skill (e.g., the request corresponds to a control of a device process, e.g., ending the timer) and so thesecond system120bmay route (246) data related to the request to thedevice skill191bof thesecond system120b. Such data may include NLU results data (such asNLU output data885/rankedoutput data825 as discussed below) and/or may include processed data that is based on such NLU results data but has been converted into a particular command specific for the device process.

In the example of the utterance of “Kitchen, turn off the timer,”second system120bmay usestate data194 to determine that adevice110 can process a <stoptimer> command and thussecond system120bmay determine that the command corresponds to a request to control a device process. Note that in this particular example, thesecond system120bmay not have information that indicates a timer is even active with regard to thefirst system120aor which assistant controls that timer. Instead, thesecond system120bmay have interpreted the command to end a timer, determined that such a command relates to a device process, and thus routed processing of the command to thedevice skill191b(and ultimately to the multi-assistant component115), even if thesecond system120bmay not have had information about the original timer. In some implementations, thedevice skill191bselected is specific to processes managed or arbitrated by themulti-assistant component115.

Thesecond system120bmay thus invoke the device skill. Thesecond system120bmay send (246) to thedevice skill191bdata representing the request, for example an instruction to send a timer, or the like. Thedevice skill191bmay then generate (248) output data to be included in a message to the device. Thedevice skill191bmay send (250) the message data back to thedevice110. Such a message may be routed through thesecond assistant component140b. The message data may include a directive to control a device process, in this example a directive to stop the timer. Thesecond assistant component140bmay receive the message data and may determine that it includes a directive corresponding to a device process control (e.g., stopping a timer). The message data may also (or in the alternative) include information indicating that the message corresponds to a process to be managed by themulti-assistant component115. Thesecond assistant component140bmay process the message data to determine (based on the directive and/or routing instructions) that the request should be handled by themulti-assistant component115. Thesecond assistant component140bmay then route the message/directive data (or a portion thereof) (253) to themulti-assistant component115. Thus, as the message data may be coming from a component (thedevice skill191b) associated with thesecond system120bit may be first routed through thesecond assistant component140bbefore being sent to themulti-assistant component115.

The message/output data may include the utterance identifier that links the particular output data to the input data (e.g., the audio data232). In this manner, thesecond system120bmay indicate to thedevice110 that the particular message/output data goes with a particular input that thedevice110 received. In this example, the message/output data may include a directive/command to thedevice110 to stop a timer.

The message data may also include data to be output as an acknowledgement of the command, such as synthesized speech acknowledging the command, data to be shown on a display of thedevice110 acknowledging the command, or the like. If such acknowledgement is included, it may be output by the device, for example by sending (252) output audio of the acknowledgement message from thesecond assistant component140bto thespeaker112 for output as synthesized speech (e.g., “stopping the timer”). The output acknowledgement message to the user may also be visual, such as a display element indicating the timer is being cancelled.

Themulti-assistant component115 may then take actions to execute the directive that was received from thedevice skill191b. In the above example, this may include executing the request to terminate a timer. As the request to stop the timer came from thesecond system120b, themessage data250 may not indicate which timer is to be stopped. Themulti-assistant component115 may take further actions to determine which timer is to be stopped.

Specifically, themulti-assistant component115 may receive the message data/directive and determine the directive relates to controlling a specific device process, namely controlling the timer. Themulti-assistant component115 may evaluate itsstate data194mto determine timer information. Such information may be available in thestate data194mas a result of thefirst assistant component140aregistering the timer information with the multi-assistant component115 (for example in relation to the operations ofFIG.2A). By evaluating thestate data194mthemulti-assistant component115 may determine that a timer is associated with the device110 (and/or user profile of the device) is active, timer controls for the device, assistant corresponding to the timer (e.g., thefirst assistant component140a), etc. Thus, by referring to thestate data194m, themulti-assistant component115 may determine that the command to end the timer corresponds to a timer that was originally initiated using thefirst system120a. Themulti-assistant component115 may then coordinate with thefirst assistant component140ato execute (254) the directive to end the timer. Specifically, themulti-assistant component115 may send thefirst assistant component140aan instruction to end the timer. Thefirst assistant component140amay then take steps to end the timer as it normally would (for example if a user pressed a button to end the timer, the command to end the timer came from thefirst system120a, etc.). Thefirst assistant component140amay also update (256) itsown state data194ato indicate information about the stopped timer and may send (258) a message to thefirst system120ato advise that the requested timer has been stopped. Thefirst system120amay then update (260) its own state data (e.g.,system state data195a) to indicate the requested timer has stopped. Themulti-assistant component115 may also update (255) the device/userprofile state data194mto note the timer has stopped.

Themulti-assistant component115 may thus coordinate with other components of thedevice110 to stop the timer, which may include ceasing output of the audio (e.g., beeping) associated with the timer.

The techniques illustrated above with regard toFIGS.2A and2B may also be implemented for many other device processes, with such cross-assistant device controls being coordinated by themulti-assistant component115 and the device skill(s)191. These configurations and techniques allow commands intended for processing using one assistant to be successfully processed by the overall system, even when the command invokes the incorrect assistant, without sharing information directly between assistants/speech processing systems.

Using an example along the lines of those discussed above, if a user initiates a process using a first speech-processing system120a(such as described above usingFIG.2A) but then speaks a command of “stop” while invoking a second speech-processing system120b, the device/profile state data194 used to process the command of “stop” may indicate what processes are active/what “stop”-type commands are available for the particular device/user profile. Thus the second speech-processing system120bmay use the device/profile state data194 to interpret the command as relating to one of the available stop commands, route the processing through adevice skill191band eventually back to themulti-assistant component115, so themulti-assistant component115 can coordinate with thefirst assistant140ato ultimately stop the active process that was initiated using the first speech-processing system120a.

In certain configurations, the device that captures the audio of the utterance may be different than the actual device to be controlled. For example, a user may speak an utterance to asmart watch110c(shown inFIG.12) along the lines of “Alexa, stop the music.” The utterance may refer to music playing not on thesmart watch110c, but on another device, for example, on a home audio system. The speech-processing system that receives the audio data of the utterance (for example,first system120aassociated with the wakeword “Alexa”) may, as described herein, determine the utterance corresponds to a device control (for example, usinglanguage processing components192a) and may send data corresponding to the utterance to thedevice skill191a. Thefirst system120amay also determine that the utterance originated from a device/user associated with a particular user profile. Thefirst system120amay send an indication of that user profile to thedevice skill191a. Alternatively (or in addition) thedevice skill191amay determine that the utterance originated from a particular device (e.g.,smart watch110c) or user associated with a specific user profile. Thefirst system120aand/ordevice skill191amay determine that the user profile is also associated with anaudio output device110athat is capable of playing music, is playing music, and/or is generally capable of performing music control operation(s) (for example as indicated bystate data194 of the user profile) regardless of whether such music playback was initiated as a result of a command tofirst system120aor to a different system such assecond system120b. Thedevice skill191amay then determine output data, including a command to stop music playback, and may send that output data to theaudio output device110athrough that device'smulti-assistant component115, for example using a process as illustrated above inFIG.2B. Thedevice skill191amay also send other output data to thesmart watch110cto acknowledge receipt and/or handling of the request to stop the music playback.

FIG.3A shows a flowchart illustrating example operations of a system for performing cross-assistant command processing. As shown, asystem120 may receive (302) audio data representing an utterance from adevice110 capable of interacting with multiple different assistants/speech-processing systems. Thesystem120 may receive (304) state data corresponding to device process controls capable of being performed either by thedevice110 or by another device referred to in the utterance (e.g., another device associated with a same user profile as device110). Such state data (e.g., device/profile state data194) may be received from thedevice110 or from some other data source (e.g.,profile storage470 discussed below,device skill191a, and/or other source). Thesystem120 may then perform (306) speech processing on the audio data to determine NLU results data. Such speech processing may use thestate data194 received above. The system may also use thestate data194 to determine (308) that the NLU results data corresponds to a request to control a device process. As a result, the system may send (310) data representing the request to a device control skill191. The data may include the NLU results data or some other version of data indicating the device process to be controlled, the control to be executed, the device to be controlled, a user profile or other identifier corresponding to the request, and/or other information related to the request. The device skill191 and/or other component may then send (312) output data to the device110 (e.g., to the multi-assistant component115). The output data may include a command to control the process and/or other data, such as acknowledgement data discussed above. Execution of the related command/directive may be coordinated betweenmulti-assistant component115 and an assistant component140. Other details of these operations are discussed herein.

FIG.3B shows a flowchart illustrating example operations of a device for performing cross-assistant command processing. As shown, adevice110 may capture (322) an utterance requesting control of a device process. Thedevice110 may determine (324) that the utterance corresponds to an invocation of a first speech-processing system. Such invocation may be the result of detection of a particular wakeword of thefirst system120a, a press of a button corresponding to thefirst system120a, detection of a gesture associated with thefirst system120a, interaction with a device corresponding tofirst system120a, or the like. Thedevice110 may then send (326) audio data representing the utterance from afirst assistant component140ato thefirst system120a. Thedevice110 may also send (328)state data194 to thefirst system120a.Such state data194 may be sent by the multi-assistant component115 (which may use thefirst assistant component140aas an intermediary). Thestate data194 may indicate a capability of thedevice110 but may not indicate the system that was originally invoked to initiate the process to be controlled. Thedevice110 may then receive (330) from thefirst system120a(e.g., from adevice skill191a) to amulti-assistant component115, a command to control the device process. Themulti-assistant component115 may receive the command and cause the device or an associated assistant component140 to coordinate (332) control of the requested device process (e.g., adjusting a volume, stopping music playback, controlling a timer, etc.). As noted above, themulti-assistant component115 may coordinate with an assistant component(s)140 to execute device control/a specific directive. Themulti-assistant component115 may determine (334) that the device process related to asecond system120b. For example, themulti-assistant component115 may evaluatestate data194m, or other data, indicating the source of the original command. Themulti-assistant component115 may then send (336) an indication of the process control to thesecond assistant component140bwhich may then send (338) the indication to thesecond system120b, thus allowing thesecond system120bto update itsown state data195bto track the control (e.g., the termination, adjustment, etc.) of the process that was originally initiated as a result of a command involvingsecond system120b.

Thesystem100 may operate using various components as described inFIG.4. The various components may be located on same or different physical devices. Communication between various components may occur directly or across a network(s)199. Thedevice110 may include audio capture component(s), such as a microphone or array of microphones of adevice110, capturesaudio11 and creates corresponding audio data. Once speech is detected in audio data representing the audio11, thedevice110 may determine if the speech is directed at thedevice110/system120. In at least some embodiments, such determination may be made using awakeword detection component420. Thewakeword detection component420 may be configured to detect various wakewords. In at least some examples, a wakeword may correspond to a name of a different digital assistant. An example wakeword/digital assistant name is “Alexa.” In another example, input to the system may be in form oftext data413, for example as a result of a user typing an input into a user interface ofdevice110. Other input forms may include indication that the user has pressed a physical or virtual button ondevice110, the user has made a gesture, etc. Thedevice110 may also capture images using camera(s)1018 of thedevice110 and may sendimage data421 representing those image(s) to thesystem120. Theimage data421 may include raw image data or image data processed by thedevice110 before sending to thesystem120.

Thewakeword detector420 of thedevice110 may process the audio data, representing the audio11, to determine whether speech is represented therein. Thedevice110 may use various techniques to determine whether the audio data includes speech. In some examples, thedevice110 may apply voice-activity detection (VAD) techniques. Such techniques may determine whether speech is present in audio data based on various quantitative aspects of the audio data, such as the spectral slope between one or more frames of the audio data; the energy levels of the audio data in one or more spectral bands; the signal-to-noise ratios of the audio data in one or more spectral bands; or other quantitative aspects. In other examples, thedevice110 may implement a classifier configured to distinguish speech from background noise. The classifier may be implemented by techniques such as linear classifiers, support vector machines, and decision trees. In still other examples, thedevice110 may apply hidden Markov model (HMM) or Gaussian mixture model (GMM) techniques to compare the audio data to one or more acoustic models in storage, which acoustic models may include models corresponding to speech, noise (e.g., environmental noise or background noise), or silence. Still other techniques may be used to determine whether speech is present in audio data.

Wakeword detection is may be performed without performing linguistic analysis, textual analysis, or semantic analysis. Instead, the audio data, representing the audio11, is analyzed to determine if specific characteristics of the audio data match preconfigured acoustic waveforms, audio signatures, or other data corresponding to a wakeword.

Thus, thewakeword detection component420 may compare audio data to stored data to detect a wakeword. One approach for wakeword detection applies general large vocabulary continuous speech recognition (LVCSR) systems to decode audio signals, with wakeword searching being conducted in the resulting lattices or confusion networks. Another approach for wakeword detection builds HMMs for each wakeword and non-wakeword speech signals, respectively. The non-wakeword speech includes other spoken words, background noise, etc. There can be one or more HMMs built to model the non-wakeword speech characteristics, which are named filler models. Viterbi decoding is used to search the best path in the decoding graph, and the decoding output is further processed to make the decision on wakeword presence. This approach can be extended to include discriminative information by incorporating a hybrid DNN-HMM decoding framework. In another example, thewakeword detection component420 may be built on deep neural network (DNN)/recursive neural network (RNN) structures directly, without HMM being involved. Such an architecture may estimate the posteriors of wakewords with context data, either by stacking frames within a context window for DNN, or using RNN. Follow-on posterior threshold tuning or smoothing is applied for decision making. Other techniques for wakeword detection, such as those known in the art, may also be used.

Once the wakeword is detected by thewakeword detector420 and/or input is detected by an input detector, thedevice110 may “wake” and begin transmittingaudio data411, representing the audio11, to the system(s)120. Theaudio data411 may include data corresponding to the wakeword; in other embodiments, the portion of the audio corresponding to the wakeword is removed by thedevice110 prior to sending theaudio data411 to the system(s)120. In the case of touch input detection or gesture based input detection, the audio data may not include a wakeword.

In some implementations, thesystem100 may include more than onesystem120. Thesystems120 may respond to different wakewords and/or perform different categories of tasks. Asystem120 may be associated with its own wakeword such that speaking a certain wakeword results in audio data be sent to and processed by a particular system. For example, detection of the wakeword “Alexa” by thewakeword detector420 may result in sending audio data tosystem120afor processing while detection of the wakeword “Mandy” by the wakeword detector may result in sending audio data tosystem120bfor processing. The system may have a separate wakeword and system for different skills/systems (e.g., “Dungeon Master” for a game play skill/system120c) and/or such skills/systems may be coordinated by one or more skill(s)490 of one ormore systems120.

Upon receipt by the system(s)120, theaudio data411 may be sent to anorchestrator component430. Theorchestrator component430 may include memory and logic that enables theorchestrator component430 to transmit various pieces and forms of data to various components of the system, as well as perform other operations as described herein.

Theorchestrator component430 may send theaudio data411 to a language processing component192. The language processing component192 (sometimes also referred to as a spoken language understanding (SLU) component) includes an automatic speech recognition (ASR)component450 and a natural language understanding (NLU)component460. TheASR component450 may transcribe theaudio data411 into text data. The text data output by theASR component450 represents one or more than one (e.g., in the form of an N-best list) ASR hypotheses representing speech represented in theaudio data411. TheASR component450 interprets the speech in theaudio data411 based on a similarity between theaudio data411 and pre-established language models. For example, theASR component450 may compare theaudio data411 with models for sounds (e.g., acoustic units such as phonemes, senons, phones, etc.) and sequences of sounds to identify words that match the sequence of sounds of the speech represented in theaudio data411. TheASR component450 sends the text data generated thereby to anNLU component460, via, in some embodiments, theorchestrator component430. The text data sent from theASR component450 to theNLU component460 may include a single top-scoring ASR hypothesis or may include an N-best list including multiple top-scoring ASR hypotheses. An N-best list may additionally include a respective score associated with each ASR hypothesis represented therein. TheASR component450 is described in greater detail below with regard toFIG.6.

The speech processing system192 may further include aNLU component460. TheNLU component460 may receive the text data from the ASR component. TheNLU component460 may attempts to make a semantic interpretation of the phrase(s) or statement(s) represented in the text data input therein by determining one or more meanings associated with the phrase(s) or statement(s) represented in the text data. TheNLU component460 may determine an intent representing an action that a user desires be performed and may determine information that allows a device (e.g., thedevice110, the system(s)120, a skill component490, skill processing component(s)125, etc.) to execute the intent. For example, if the text data corresponds to “play the 5^thSymphony by Beethoven,” theNLU component460 may determine an intent that the system output music and may identify “Beethoven” as an artist/composer and “5th Symphony” as the piece of music to be played. For further example, if the text data corresponds to “what is the weather,” theNLU component460 may determine an intent that the system output weather information associated with a geographic location of thedevice110. In another example, if the text data corresponds to “turn off the lights,” theNLU component460 may determine an intent that the system turn off lights associated with thedevice110 or the user5.

As noted above, in certain instances a user may issue a request to control a device process. In such a situation, the user may speak the request to one assistant system120 (for example asecond system120b) to control a device process that was actually initiated with a command to another assistant system (for example afirst system120a). To correctly process such a request without thesecond system120bhaving information about processes related to thefirst system120a, theNLU460 may have access tostate data194 that allows asystem120 to determine that the request corresponds to a device control that is capable of being executed by thedevice110, through an interface with device skill191.

TheNLU component460 may returnNLU results data885/825 (which is further discussed below in reference toFIG.8 and may include tagged text data, indicators of intent, etc.) back to theorchestrator430. Theorchestrator430 may forward the NLU results data to a skill component(s)490. If the NLU results data includes a single NLU hypothesis, theNLU component460 and theorchestrator component430 may direct the NLU results data to the skill component(s)490 associated with the NLU hypothesis. If theNLU results data885/825 includes an N-best list of NLU hypotheses, theNLU component460 and theorchestrator component430 may direct the top scoring NLU hypothesis to a skill component(s)490 associated with the top scoring NLU hypothesis. The system may also include apost-NLU ranker465 which may incorporate other information to rank potential interpretations determined by theNLU component460. Thelocal device110 may also include its own post-NLU ranker, which may operate similarly to thepost-NLU ranker465. TheNLU component460,post-NLU ranker465 and other components are described in greater detail below with regard toFIGS.7 and8.

A skill component may be software running on the system(s)120 that is akin to a software application. That is, a skill component490 may enable the system(s)120 to execute specific functionality in order to provide data or produce some other requested output. As used herein, a “skill component” may refer to software that may be placed on a machine or a virtual machine (e.g., software that may be launched in a virtual instance when called). A skill component may be software customized to perform one or more actions as indicated by a business entity, device manufacturer, user, etc. What is described herein as a skill component may be referred to using many different terms, such as an action, bot, app, or the like. The system(s)120 may be configured with more than one skill component490. For example, a weather service skill component may enable the system(s)120 to provide weather information, a car service skill component may enable the system(s)120 to book a trip with respect to a taxi or ride sharing service, a restaurant skill component may enable the system(s)120 to order a pizza with respect to the restaurant's online ordering system, etc. A skill component490 may operate in conjunction between the system(s)120 and other devices, such as thedevice110, in order to complete certain functions. Inputs to a skill component490 may come from speech processing interactions or through other interactions or input sources. A skill component490 may include hardware, software, firmware, or the like that may be dedicated to a particular skill component490 or shared among different skill components490.

Skill processing component(s)125 may communicate with a skill component(s)490 within the system(s)120 and/or directly with theorchestrator component430 or with other components. A skill processing component(s)125 may be configured to perform one or more actions. An ability to perform such action(s) may sometimes be referred to as a “skill.” That is, a skill may enable a skill processing component(s)125 to execute specific functionality in order to provide data or perform some other action requested by a user. For example, a weather service skill may enable a skill processing component(s)125 to provide weather information to the system(s)120, a car service skill may enable a skill processing component(s)125 to book a trip with respect to a taxi or ride sharing service, an order pizza skill may enable a skill processing component(s)125 to order a pizza with respect to a restaurant's online ordering system, etc. Additional types of skills include home automation skills (e.g., skills that enable a user to control home devices such as lights, door locks, cameras, thermostats, etc.), entertainment device skills (e.g., skills that enable a user to control entertainment devices such as smart televisions), video skills, flash briefing skills, as well as custom skills that are not associated with any pre-configured type of skill.

The system(s)120 may be configured with a skill component490 dedicated to interacting with the skill processing component(s)125. Unless expressly stated otherwise, reference to a skill, skill device, or skill component may include a skill component490 operated by the system(s)120 and/or skill operated by the skill processing component(s)125. Moreover, the functionality described herein as a skill or skill may be referred to using many different terms, such as an action, bot, app, or the like. The skill490 and or skill processing component(s)125 may return output data to theorchestrator430.

Dialog processing is a field of computer science that involves communication between a computing system and a human via text, audio, and/or other forms of communication. While some dialog processing involves only simple generation of a response given only a most recent input from a user (i.e., single-turn dialog), more complicated dialog processing involves determining and optionally acting on one or more goals expressed by the user over multiple turns of dialog, such as making a restaurant reservation and/or booking an airline ticket. These multi-turn “goal-oriented” dialog systems may recognize, retain, and use information collected during more than one input during a back-and-forth or “multi-turn” interaction with the user; for example, information regarding a language in which a dialog is being conducted.

The system(s)100 may include adialog manager component572 that manages and/or tracks a dialog between a user and a device, and in some cases between the user and one ormore systems120. As used herein, a “dialog” may refer to data transmissions (such as relating to multiple user inputs andsystem100 outputs) between thesystem100 and a user (e.g., through device(s)110) that all relate to a single “conversation” between the system and the user that may have originated with a single user input initiating the dialog. Thus, the data transmissions of a dialog may be associated with a same dialog identifier, which may be used by components of theoverall system100 to track information across the dialog. Subsequent user inputs of the same dialog may or may not start with speaking of a wakeword. Each natural language input of a dialog may be associated with a different natural language input identifier such that multiple natural language input identifiers may be associated with a single dialog identifier. Further, other non-natural language inputs (e.g., image data, gestures, button presses, etc.) may relate to a particular dialog depending on the context of the inputs. For example, a user may open a dialog with thesystem100 to request a food delivery in a spoken utterance and the system may respond by displaying images of food available for order and the user may speak a response (e.g., “item1” or “that one”) or may gesture a response (e.g., point to an item on the screen or give a thumbs-up) or may touch the screen on the desired item to be selected. Non-speech inputs (e.g., gestures, screen touches, etc.) may be part of the dialog and the data associated therewith may be associated with the dialog identifier of the dialog.

Thedialog manager component572 may associate a dialog session identifier with the dialog upon identifying that the user is engaging in a dialog with the user. Thedialog manager component572 may track a user input and the corresponding system generated response to the user input as a turn. The dialog session identifier may correspond to multiple turns of user input and corresponding system generated response. Thedialog manager component572 may transmit data identified by the dialog session identifier directly to theorchestrator component430 or other component. Depending on system configuration thedialog manager572 may determine the appropriate system generated response to give to a particular utterance or user input of a turn. Or creation of the system generated response may be managed by another component of the system (e.g., the language output component193,NLG479,orchestrator430, etc.) while thedialog manager572 selects the appropriate responses. Alternatively, another component of the system(s)120 may select responses using techniques discussed herein. The text of a system generated response may be sent to aTTS component480 for creation of audio data corresponding to the response. The audio data may then be sent to a user device (e.g., device110) for ultimate output to the user. Alternatively (or in addition) a dialog response may be returned in text or some other form.

Thedialog manager572 may receive the ASR hypothesis/hypotheses (i.e., text data) and make a semantic interpretation of the phrase(s) or statement(s) represented therein. That is, thedialog manager572 determines one or more meanings associated with the phrase(s) or statement(s) represented in the text data based on words represented in the text data. Thedialog manager572 determines a goal corresponding to an action that a user desires be performed as well as pieces of the text data that allow a device (e.g., thedevice110, the system(s)120, a skill490, a skill processing component(s)125, etc.) to execute the intent. If, for example, the text data corresponds to “what is the weather,” thedialog manager572 may determine that that the system(s)120 is to output weather information associated with a geographic location of thedevice110. In another example, if the text data corresponds to “turn off the lights,” thedialog manager572 may determine that the system(s)120 is to turn off lights associated with the device(s)110 or the user(s)5.

Thedialog manager572 may send the results data to one or more skill(s)490. If the results data includes a single hypothesis, theorchestrator component430 may send the results data to the skill(s)490 associated with the hypothesis. If the results data includes an N-best list of hypotheses, theorchestrator component430 may send the top scoring hypothesis to a skill(s)490 associated with the top scoring hypothesis.

Thesystem120 includes a language output component193. The language output component193 includes a natural language generation (NLG)component479 and a text-to-speech (TTS)component480. TheNLG component479 can generate text for purposes of TTS output to a user. For example theNLG component479 may generate text corresponding to instructions corresponding to a particular action for the user to perform. TheNLG component479 may generate appropriate text for various outputs as described herein. TheNLG component479 may include one or more trained models configured to output text appropriate for a particular input. The text output by theNLG component479 may become input for the TTS component480 (e.g., output text data1010 discussed below). Alternatively or in addition, theTTS component480 may receive text data from a skill490 or other system component for output.

TheNLG component479 may include a trained model. TheNLG component479 generates text data1010 from dialog data received by thedialog manager572 such that the output text data1010 has a natural feel and, in some embodiments, includes words and/or phrases specifically formatted for a requesting individual. The NLG may use templates to formulate responses. And/or the NLG system may include models trained from the various templates for forming the output text data1010. For example, the NLG system may analyze transcripts of local news programs, television shows, sporting events, or any other media program to obtain common components of a relevant language and/or region. As one illustrative example, the NLG system may analyze a transcription of a regional sports program to determine commonly used words or phrases for describing scores or other sporting news for a particular region. The NLG may further receive, as inputs, a dialog history, an indicator of a level of formality, and/or a command history or other user history such as the dialog history.

The NLG system may generate dialog data based on one or more response templates. Further continuing the example above, the NLG system may select a template in response to the question, “What is the weather currently like?” of the form: “The weather currently is $weather_information$.” The NLG system may analyze the logical form of the template to produce one or more textual responses including markups and annotations to familiarize the response that is generated. In some embodiments, the NLG system may determine which response is the most appropriate response to be selected. The selection may, therefore, be based on past responses, past questions, a level of formality, and/or any other feature, or any other combination thereof. Responsive audio data representing the response generated by the NLG system may then be generated using the text-to-speech component480.

TheTTS component480 may generate audio data (e.g., synthesized speech) from text data using one or more different methods. Text data input to theTTS component480 may come from a skill component490, theorchestrator component430, or another component of the system. In one method of synthesis called unit selection, theTTS component480 matches text data against a database of recorded speech. TheTTS component480 selects matching units of recorded speech and concatenates the units together to form audio data. In another method of synthesis called parametric synthesis, theTTS component480 varies parameters such as frequency, volume, and noise to create audio data including an artificial speech waveform. Parametric synthesis uses a computerized voice generator, sometimes called a vocoder. TheTTS component480 may be capable of generating output audio representing natural language speech in one or more natural languages (e.g., English, Mandarin, French, etc.).

The system100 (either ondevice110,system120, or a combination thereof) may include profile storage for storing a variety of information related to individual users, groups of users, devices, etc. that interact with the system. As used herein, a “profile” refers to a set of data associated with a user, group of users, device, etc. The data of a profile may include preferences specific to the user, device, etc.; input and output capabilities of the device; internet connectivity information; user bibliographic information; subscription information, as well as other information.

Theprofile storage470 may include one or more user profiles, with each user profile being associated with a different user identifier/user profile identifier. Each user profile may include various user identifying data. Each user profile may also include data corresponding to preferences of the user. Each user profile may also include preferences of the user and/or one or more device identifiers, representing one or more devices of the user. For instance, the user account may include one or more IP addresses, MAC addresses, and/or device identifiers, such as a serial number, of each additional electronic device associated with the identified user account. When a user logs into to an application installed on adevice110, the user profile (associated with the presented login information) may be updated to include information about thedevice110, for example with an indication that the device is currently in use. Each user profile may include identifiers of skills that the user has enabled. When a user enables a skill, the user may give thesystem120 permission to allow the skill to execute with respect to the user's natural language user inputs. If a user does not enable a skill, thesystem120 may not invoke the skill to execute with respect to the user's natural language user inputs.

Theprofile storage470 may include one or more group profiles. Each group profile may be associated with a different group identifier. A group profile may be specific to a group of users. That is, a group profile may be associated with two or more individual user profiles. For example, a group profile may be a household profile that is associated with user profiles associated with multiple users of a single household. A group profile may include preferences shared by all the user profiles associated therewith. Each user profile associated with a group profile may additionally include preferences specific to the user associated therewith. That is, each user profile may include preferences unique from one or more other user profiles associated with the same group profile. A user profile may be a stand-alone profile or may be associated with a group profile.

Theprofile storage470 may include one or more device profiles. Each device profile may be associated with a different device identifier. Each device profile may include various device identifying information. Each device profile may also include one or more user identifiers, representing one or more users associated with the device. For example, a household device's profile may include the user identifiers of users of the household.

Theprofile storage470 may include data corresponding tostate data194. For example, theprofile storage470 may indicate the device process control capabilities of one ormore devices110 associated with a particular user profile.Such state data194 may be updated by one or more device(s)110 as user(s) interact with the device(s) to maintain an updated record of the state of the device. Alternatively (or in addition) theprofile storage470 may include, for a particular user profile,state data194 reflecting capability data indicating the device process control operations that may be performed by adevice110.

Although the components ofFIG.4 may be illustrated as part of system(s)120,device110, or otherwise, the components may be arranged in other device(s) (such as indevice110 if illustrated in system(s)120 or vice-versa, or in other device(s) altogether) without departing from the disclosure.FIG.5 illustrates such a configureddevice110.

In theexample system100 shown inFIG.5, thedevice110 includes thefirst assistant component140aand thesecond assistant component140b. Thefirst assistant component140amay be in communication with back-end components of thefirst system120a(e.g., via the network199). Thefirst assistant component140amay also be in communication with thelanguage processing component592, thelanguage output component593, afirst wakeword detector121a, and/orhybrid selector524. Thefirst system120amay be associated with one or morelocal skill components190a1,190a2, and190a3 (collectively “skill components190”). The local skill components190 may be in communication with one or more skill processing component(s)125. Thesecond assistant component140bmay be associated with thesecond system120b, which may be a separate computing system separate and remote from thedevice110. Thefirst system120aand thesecond system120bbe configured as described herein; for example, as described with respect toFIG.1 andFIG.4.

Thesecond assistant component140bmay be logically or otherwise walled off from certain components of thedevice110. For example thesecond assistant component140bmay not be able to communicate directly with thefirst assistant component140a; such communications may need to be mediated bymulti-assistant component115. Thesecond assistant component140bmay include or be associated with its own proprietary components. For example, thesecond assistant component140bmay be associated with asecond wakeword detector121b. In addition, thesecond assistant component140bmay leverage separate language processing and language output components, which may reside in thedevice110 or thesecond system120b. Thesecond assistant component140bmay, however, interface with amulti-assistant component115 and/or adialog manager472, which may be shared between thefirst assistant component140aand thesecond assistant component140b.

In some implementations, speech processing of input audio data directed to thefirst system120amay take place on thedevice110. Thedevice110 may send a message represented in the input audio data to thesecond system120bwithout first sending the input audio data to thefirst system120a. For example, thedevice110 may receive the input audio data and detect, with the firstwakeword detection component121a, a wakeword corresponding to thefirst system120a. Thelanguage processing components592 of thedevice110 may process the input audio data and determine that the input audio data represents a request to generate a message and send the message to thesecond system120b. Thefirst assistant component140amay receive the output of thelanguage processing components592, and forward it to themulti-assistant component115. Thefirst assistant component140amay include with the output metadata that indicates that themulti-assistant component115 is to forward the output to thesecond system120b(e.g., via thesecond assistant component140b). In some cases, thefirst assistant component140amay send the output to thelanguage output components593 to generate an output in the form of output audio data (e.g., a TTS output) representing the output. Themulti-assistant component593 may receive the output (or output audio data) and metadata, and determine that the output is to be processed by thesecond system120b. Themulti-assistant component115 may send the output to thesecond assistant component140b. Thesecond assistant component140bmay send the output to thesecond system120b. Thesecond system120bmay process the output by, for example, executing a command represented in the output. Thesystem120bmay return response data to thedevice110; for example, by sending responsive output audio data to themulti-assistant component115 for output by a speaker of the device.

In some cases, themulti-assistant component115 may determine (for example, based on state data regarding an active dialog that includes the input audio data) that the response data from thesecond system120bis to be translated back into the language of the input audio data. Themulti-assistant component115 may send the response data to thefirst system120avia thefirst assistant component140aalong with an indication that the response data is to be translated. The response data may, for example, be audio data and/or text data. Thefirst system120amay return translated response data. The translated response data may be audio data and/or text data, if the translated response data is text data, themulti-assistant component115 may send it to thelanguage output components593 for conversion into synthetic speech for output by thedevice110.

In at least some embodiments, thesystem120 may receive theaudio data411 from thedevice110, to recognize speech corresponding to a spoken input in the receivedaudio data411, and to perform functions in response to the recognized speech. In at least some embodiments, these functions involve sending directives (e.g., commands), from thesystem120 to the device110 (and/or other devices110) to cause thedevice110 to perform an action, such as output an audible response to the spoken input via a loudspeaker(s), and/or control secondary devices in the environment by sending a control command to the secondary devices.

Thus, when thedevice110 is able to communicate with thesystem120 over the network(s)199, some or all of the functions capable of being performed by thesystem120 may be performed by sending one or more directives over the network(s)199 to thedevice110, which, in turn, may process the directive(s) and perform one or more corresponding actions. For example, thesystem120, using a remote directive that is included in response data (e.g., a remote response), may instruct thedevice110 to output an audible response (e.g., using TTS processing performed by an on-device TTS component580) to a user's question via a loudspeaker(s) of (or otherwise associated with) thedevice110, to output content (e.g., music) via the loudspeaker(s) of (or otherwise associated with) thedevice110, to display content on a display of (or otherwise associated with) thedevice110, and/or to send a directive to a secondary device (e.g., a directive to turn on a smart light). It is to be appreciated that thesystem120 may be configured to provide other functions in addition to those discussed herein, such as, without limitation, providing step-by-step directions for navigating from an origin location to a destination location, conducting an electronic commerce transaction on behalf of the user5 as part of a shopping function, establishing a communication session (e.g., a video call) between the user5 and another user, and so on.

Thedevice110 may include one or more wakeword detection components121 (and/or121aand/or121b) configured to compare theaudio data411 to stored models used to detect a wakeword (e.g., “Alexa”) that indicates to thedevice110 that theaudio data411 is to be processed for determining NLU output data (e.g., slot data that corresponds to a named entity, label data, and/or intent data, etc.). In at least some embodiments, ahybrid selector524, of thedevice110, may send theaudio data411 to thewakeword detection component121a. If thewakeword detection component121adetects a wakeword in theaudio data411, thewakeword detection component121amay send an indication of such detection to thehybrid selector524. In response to receiving the indication, thehybrid selector524 may send theaudio data411 to thesystem120 and/or theASR component550. Thewakeword detection component121amay also send an indication, to thehybrid selector524, representing a wakeword was not detected. In response to receiving such an indication, thehybrid selector524 may refrain from sending theaudio data411 to thesystem120, and may prevent theASR component550 from further processing theaudio data411. In this situation, theaudio data411 can be discarded.

Thedevice110 may conduct its own speech processing using on-device language processing components, such as an SLU/language processing component592 (which may include anASR component550 and an NLU560), similar to the manner discussed herein with respect to the SLU component192 (orASR component450 and the NLU component460) of thesystem120.Language processing component592 may operate similarly to language processing component192,ASR component550 may operate similarly toASR component450 andNLU component560 may operate similarly toNLU component460. Thedevice110 may also internally include, or otherwise have access to, other components such as one or more skill components190 capable of executing commands based on NLU output data or other results determined by thedevice110/system120 (which may operate similarly to skill components490), profile storage570 (configured to store similar profile data to that discussed herein with respect to theprofile storage470 of the system120), or other components. In at least some embodiments, theprofile storage570 may only store profile data for a user or group of users specifically associated with thedevice110. Similar to as described above with respect to skill component490, a skill component190 may communicate with a skill processing component(s)125. Thedevice110 may also have its ownlanguage output component593 which may includeNLG component579 andTTS component580.Language output component593 may operate similarly to language processing component192,NLG component579 may operate similarly toNLG component479 andTTS component580 may operate similarly toTTS component480.

In at least some embodiments, the on-device language processing components may not have the same capabilities as the language processing components of thesystem120. For example, the on-device language processing components may be configured to handle only a subset of the natural language user inputs that may be handled by thesystem120. For example, such subset of natural language user inputs may correspond to local-type natural language user inputs, such as those controlling devices or components associated with a user's home. In such circumstances the on-device language processing components may be able to more quickly interpret and respond to a local-type natural language user input, for example, than processing that involves thesystem120. If thedevice110 attempts to process a natural language user input for which the on-device language processing components are not necessarily best suited, the language processing results determined by thedevice110 may indicate a low confidence or other metric indicating that the processing by thedevice110 may not be as accurate as the processing done by thesystem120.

Thehybrid selector524, of thedevice110, may include a hybrid proxy (HP)526 configured to proxy traffic to/from thesystem120. For example, theHP526 may be configured to send messages to/from a hybrid execution controller (HEC)527 of thehybrid selector524. For example, command/directive data received from thesystem120 can be sent to theHEC527 using theHP526. TheHP526 may also be configured to allow theaudio data411 to pass to thesystem120 while also receiving (e.g., intercepting) thisaudio data411 and sending theaudio data411 to theHEC527.

In at least some embodiments, thehybrid selector524 may further include a local request orchestrator (LRO)528 configured to notify theASR component550 about the availability ofnew audio data411 that represents user speech, and to otherwise initiate the operations of local language processing whennew audio data411 becomes available. In general, thehybrid selector524 may control execution of local language processing, such as by sending “execute” and “terminate” events/instructions. An “execute” event may instruct a component to continue any suspended execution (e.g., by instructing the component to execute on a previously-determined intent in order to determine a directive). Meanwhile, a “terminate” event may instruct a component to terminate further execution, such as when thedevice110 receives directive data from thesystem120 and chooses to use that remotely-determined directive data.

Thus, when theaudio data411 is received, theHP526 may allow theaudio data411 to pass through to thesystem120 and theHP526 may also input theaudio data411 to the on-device ASR component550 by routing theaudio data411 through theHEC527 of thehybrid selector524, whereby theLRO528 notifies theASR component550 of theaudio data411. At this point, thehybrid selector524 may wait for response data from either or both of thesystem120 or the local language processing components. However, the disclosure is not limited thereto, and in some examples thehybrid selector524 may send theaudio data411 only to thelocal ASR component550 without departing from the disclosure. For example, thedevice110 may process theaudio data411 locally without sending theaudio data411 to thesystem120.

Thelocal ASR component550 is configured to receive theaudio data411 from thehybrid selector524, and to recognize speech in theaudio data411, and thelocal NLU component560 is configured to determine a user intent from the recognized speech, and to determine how to act on the user intent by generating NLU output data which may include directive data (e.g., instructing a component to perform an action). Such NLU output data may take a form similar to that as determined by theNLU component460 of thesystem120. In some cases, a directive may include a description of the intent (e.g., an intent to turn off {device A}). In some cases, a directive may include (e.g., encode) an identifier of a second device(s), such as kitchen lights, and an operation to be performed at the second device(s). Directive data may be formatted using Java, such as JavaScript syntax, or JavaScript-based syntax. This may include formatting the directive using JSON. In at least some embodiments, a device-determined directive may be serialized, much like how remotely-determined directives may be serialized for transmission in data packets over the network(s)199. In at least some embodiments, a device-determined directive may be formatted as a programmatic application programming interface (API) call with a same logical operation as a remotely-determined directive. In other words, a device-determined directive may mimic a remotely-determined directive by using a same, or a similar, format as the remotely-determined directive.

An NLU hypothesis (output by the NLU component560) may be selected as usable to respond to a natural language user input, and local response data may be sent (e.g., local NLU output data, local knowledge base information, internet search results, and/or local directive data) to thehybrid selector524, such as a “ReadyToExecute” response. Thehybrid selector524 may then determine whether to use directive data from the on-device components to respond to the natural language user input, to use directive data received from thesystem120, assuming a remote response is even received (e.g., when thedevice110 is able to access thesystem120 over the network(s)199), or to determine output audio requesting additional information from the user5.

Thedevice110 and/or thesystem120 may associate a unique identifier with each natural language user input. Thedevice110 may include the unique identifier when sending theaudio data411 to thesystem120, and the response data from thesystem120 may include the unique identifier to identify which natural language user input the response data corresponds.

In at least some embodiments, thedevice110 may include, or be configured to use, one or more skill components190 that may work similarly to the skill component(s)490 implemented by thesystem120. The skill component(s)190 may correspond to one or more domains that are used in order to determine how to act on a spoken input in a particular way, such as by outputting a directive that corresponds to the determined intent, and which can be processed to implement the desired operation. The skill component(s)190 installed on thedevice110 may include, without limitation, a smart home skill component (or smart home domain) and/or a device control skill component (or device control domain) to execute in response to spoken inputs corresponding to an intent to control a second device(s) in an environment, a music skill component (or music domain) to execute in response to spoken inputs corresponding to a intent to play music, a navigation skill component (or a navigation domain) to execute in response to spoken input corresponding to an intent to get directions, a shopping skill component (or shopping domain) to execute in response to spoken inputs corresponding to an intent to buy an item from an electronic marketplace, and/or the like.

Additionally or alternatively, thedevice110 may be in communication with one or more skill processing component(s)125. For example, a skill processing component(s)125 may be located in a remote environment (e.g., separate location) such that thedevice110 may only communicate with the skill processing component(s)125 via the network(s)199. However, the disclosure is not limited thereto. For example, in at least some embodiments, a skill processing component(s)125 may be configured in a local environment (e.g., home server and/or the like) such that thedevice110 may communicate with the skill processing component(s)125 via a private network, such as a local area network (LAN).

As used herein, a “skill” may refer to a skill component190/490, a skill processing component(s)125, or a combination of a skill component190/490 and a corresponding skill processing component(s)125. Similar to the manner discussed herein, thelocal device110 may be configured to recognize multiple different wakewords and/or perform different categories of tasks depending on the wakeword. Such different wakewords may invoke different processing components of local device110 (not illustrated inFIG.5). For example, detection of the wakeword “Alexa” by thewakeword detector121amay result in sending audio data to certainlanguage processing components592/skills190 for processing while detection of the wakeword “Computer” by the wakeword detector may result in sending audio data differentlanguage processing components592/skills190 for processing.

FIG.6 is a conceptual diagram of anASR component450, according to embodiments of the present disclosure. TheASR component450 may receiveaudio data631 and process it to recognize and transcribe speech contained therein. TheASR component450 may output the transcript asASR output data615. In some cases, theASR component450 may generate more than one ASR hypothesis (e.g., representing a possible transcript) for a single spoken natural language input. An ASR hypothesis may be assigned a score (e.g., probability score, confidence score, etc.) representing a likelihood that the corresponding ASR hypothesis matches the spoken natural language input (e.g., representing a likelihood that a particular set of words matches those spoken in the natural language input). The score may be based on a number of factors including, for example, a similarity of the sound in the spoken natural language input to models for language sounds (e.g., anacoustic model653 stored in the ASR model storage652), and the likelihood that a particular word, which matches the sounds, would be included in the sentence at the specific location (e.g., using a language or grammar model654). Based on the considered factors and the assigned confidence score, theASR component450 may output an ASR hypothesis that most likely matches the spoken natural language input, or may output multiple ASR hypotheses in the form of a lattice or an N-best list, with each ASR hypothesis corresponding to a respective score.

TheASR component450 may interpret a spoken natural language input using one or more models in theASR model storage652. Such models may consist of NN-based end-to-end models such as theASR model650. Some models may process theaudio data631 based on the similarity between the spoken natural language input and acoustic units (e.g., representing subword units or phonemes) in anacoustic model653, and use alanguage models654 to predict words/phrases/sentences likely represented by sequences of the acoustic units. In some implementations, a finite state transducer (FST)655 may perform language model functions.

TheASR component450 may include aspeech recognition engine658. TheASR component450 may receiveaudio data631 from, for example, amicrophone114 of auser device110. In some cases, theaudio data631 may have been processed audio detected by an acoustic front end (AFE) or other component. Thespeech recognition engine658 may process theaudio data631 using one or more of theASR model650,acoustic models653,language models654, FST(s)655, and/or other data models and information for recognizing the speech conveyed in the audio data. Theaudio data631 may be audio data that has been digitized (for example by the AFE) into frames representing time intervals for which the AFE determines a number of values, called features, representing the qualities of the audio data, along with a set of those values, called a feature vector, representing the features/qualities of the audio data within the frame. In at least some embodiments, audio frames may be 10 ms each. In some embodiments, an audio frame may represent a larger window of audio; for example, ˜2 ms. Many different features may be determined, as known in the art, and each feature may represent some quality of the audio that may be useful for ASR processing. A number of approaches may be used by an AFE to process the audio data, such as log-filterbank energies (LFBE), Mel-frequency cepstral coefficients (MFCCs), perceptual linear predictive (PLP) techniques, neural network feature vector techniques, linear discriminant analysis, semi-tied covariance matrices, or other approaches known to those of skill in the art. In some cases, feature vectors of theaudio data631 may arrive at theprocessing system120 encoded, in which case they may be decoded by thespeech recognition engine658 and/or prior to processing by thespeech recognition engine658.

In some implementations, theASR component450 may process theaudio data631 using theASR model650. TheASR model650 may be, for example, a recurrent neural network such as an RNN-T. TheASR model650 may predict a probability (y|x) of labels y=(y₁, . . . , y_u) given acoustic features x=(x₁, . . . , x_t). During inference, theASR model650 can generate an N-best list using, for example, a beam search decoding algorithm. TheASR model650 may include anencoder610, aprediction network620, ajoint network630, and asoftmax640. Theencoder610 may be similar or analogous to an acoustic model (e.g., similar to theacoustic model653 described below), and may process a sequence of acoustic input features to generate encoded hidden representations. Theprediction network620 may be similar or analogous to a language model (e.g., similar to thelanguage model654 described below), and may process the previous output label predictions, and map them to corresponding hidden representations. Thejoint network630 may be, for example, a feed forward NN that may process hidden representations from both theencoder610 andprediction network620, and predict output label probabilities. Thesoftmax640 may be a function implemented (e.g., as a layer of the joint network630) to normalize the predicted output probabilities.

In some implementations, thespeech recognition engine658 may attempt to match received feature vectors in theaudio data631 to language acoustic units (e.g., phonemes) and words as known in the storedacoustic models653,language models654, and/or FST(s)655. For example,audio data631 may be processed by one or more acoustic model(s)653 to determine acoustic unit data. The acoustic unit data may include indicators of acoustic units detected in theaudio data631 by theASR component450. For example, acoustic units can consist of one or more of phonemes, diaphonemes, tonemes, phones, diphones, triphones, or the like. The acoustic unit data can be represented using one or a series of symbols from a phonetic alphabet such as the X-SAMPA, the International Phonetic Alphabet, or Initial Teaching Alphabet (ITA) phonetic alphabets. In some implementations a phoneme representation of the audio data can be analyzed using an n-gram based tokenizer. An entity, or a slot representing one or more entities, can be represented by a series of n-grams.

The acoustic unit data may be processed using the language model654 (and/or using FST655) to determineASR output data615. TheASR output data615 can include one or more hypotheses. One or more of the hypotheses represented in theASR output data615 may then be sent to further components (such as the NLU component660) for further processing as discussed herein. TheASR output data615 may include representations of text of an utterance, such as words, subword units, or the like.

Thespeech recognition engine658 computes scores for the feature vectors based on acoustic information and language information. The acoustic information (such as identifiers for acoustic units and/or corresponding scores) is used to calculate an acoustic score representing a likelihood that the intended sound represented by a group of feature vectors matches a language phoneme. The language information is used to adjust the acoustic score by considering what sounds and/or words are used in context with each other, thereby improving the likelihood that theASR component450 will output ASR hypotheses that make sense grammatically. The specific models used may be general models or may be models corresponding to a particular domain, such as music, banking, etc.

Thespeech recognition engine658 may use a number of techniques to match feature vectors to phonemes, for example using Hidden Markov Models (HMMs) to determine probabilities that feature vectors may match phonemes. Sounds received may be represented as paths between states of the HMM and multiple paths may represent multiple possible text matches for the same sound. Further techniques, such as using FSTs, may also be used.

Thespeech recognition engine658 may use the acoustic model(s)653 to attempt to match received audio feature vectors to words or subword acoustic units. An acoustic unit may be a senone, phoneme, phoneme in context, syllable, part of a syllable, syllable in context, or any other such portion of a word. Thespeech recognition engine658 computes recognition scores for the feature vectors based on acoustic information and language information. The acoustic information is used to calculate an acoustic score representing a likelihood that the intended sound represented by a group of feature vectors match a subword unit. The language information is used to adjust the acoustic score by considering what sounds and/or words are used in context with each other, thereby improving the likelihood that theASR component450 outputs ASR hypotheses that make sense grammatically.

Thespeech recognition engine658 may use a number of techniques to match feature vectors to phonemes or other acoustic units, such as diphones, triphones, etc. One common technique is using Hidden Markov Models (HMMs). HMMs are used to determine probabilities that feature vectors may match phonemes. Using HMMs, a number of states are presented, in which the states together represent a potential phoneme (or other acoustic unit, such as a triphone) and each state is associated with a model, such as a Gaussian mixture model or a deep belief network. Transitions between states may also have an associated probability, representing a likelihood that a current state may be reached from a previous state. Sounds received may be represented as paths between states of the HMM and multiple paths may represent multiple possible text matches for the same sound. Each phoneme may be represented by multiple potential states corresponding to different known pronunciations of the phonemes and their parts (such as the beginning, middle, and end of a spoken language sound). An initial determination of a probability of a potential phoneme may be associated with one state. As new feature vectors are processed by thespeech recognition engine658, the state may change or stay the same, based on the processing of the new feature vectors. A Viterbi algorithm may be used to find the most likely sequence of states based on the processed feature vectors.

The probable phonemes and related states/state transitions, for example HMM states, may be formed into paths traversing a lattice of potential phonemes. Each path represents a progression of phonemes that potentially match the audio data represented by the feature vectors. One path may overlap with one or more other paths depending on the recognition scores calculated for each phoneme. Certain probabilities are associated with each transition from state to state. A cumulative path score may also be calculated for each path. This process of determining scores based on the feature vectors may be called acoustic modeling. When combining scores as part of the ASR processing, scores may be multiplied together (or combined in other ways) to reach a desired combined score or probabilities may be converted to the log domain and added to assist processing.

Thespeech recognition engine658 may also compute scores of branches of the paths based on language models or grammars. Language modeling involves determining scores for what words are likely to be used together to form coherent words and sentences. Application of a language model may improve the likelihood that theASR component450 correctly interprets the speech contained in the audio data. For example, for an input audio sounding like “hello,” acoustic model processing that returns the potential phoneme paths of “H E L O”, “H A L O”, and “Y E L O” may be adjusted by a language model to adjust the recognition scores of “H E L O” (interpreted as the word “hello”), “H A L O” (interpreted as the word “halo”), and “Y E L O” (interpreted as the word “yellow”) based on the language context of each word within the spoken utterance.

FIGS.7 and8 illustrates how theNLU component460 may perform NLU processing.FIG.7 is a conceptual diagram of how natural language processing is performed, according to embodiments of the present disclosure. AndFIG.8 is a conceptual diagram of how natural language processing is performed, according to embodiments of the present disclosure.

FIG.7 illustrates how NLU processing is performed on text data. TheNLU component460 may process text data including several ASR hypotheses of a single user input. For example, if theASR component450 outputs text data including an n-best list of ASR hypotheses, theNLU component460 may process the text data with respect to all (or a portion of) the ASR hypotheses represented therein.

TheNLU component460 may annotate text data by parsing and/or tagging the text data. For example, for the text data “tell me the weather for Seattle,” theNLU component460 may tag “tell me the weather for Seattle” as an <OutputWeather> intent as well as separately tag “Seattle” as a location for the weather information.

TheNLU component460 may include ashortlister component750. Theshortlister component750 selects skills that may execute with respect toASR output data615 input to the NLU component460 (e.g., applications that may execute with respect to the user input). The ASR output data615 (which may also be referred to as ASR data615) may include representations of text of an utterance, such as words, subword units, or the like. Theshortlister component750 thus limits downstream, more resource intensive NLU processes to being performed with respect to skills that may execute with respect to the user input.

Without ashortlister component750, theNLU component460 may processASR output data615 input thereto with respect to every skill of the system, either in parallel, in series, or using some combination thereof. By implementing ashortlister component750, theNLU component460 may processASR output data615 with respect to only the skills that may execute with respect to the user input. This reduces total compute power and latency attributed to NLU processing.

Theshortlister component750 may include one or more trained models. The model(s) may be trained to recognize various forms of user inputs that may be received by the system(s)120. For example, during a training period skill processing component(s)125 associated with a skill may provide the system(s)120 with training text data representing sample user inputs that may be provided by a user to invoke the skill. For example, for a ride sharing skill, a skill processing component(s)125 associated with the ride sharing skill may provide the system(s)120 with training text data including text corresponding to “get me a cab to [location],” “get me a ride to [location],” “book me a cab to [location],” “book me a ride to [location],” etc. The one or more trained models that will be used by theshortlister component750 may be trained, using the training text data representing sample user inputs, to determine other potentially related user input structures that users may try to use to invoke the particular skill. During training, the system(s)120 may solicit the skill processing component(s)125 associated with the skill regarding whether the determined other user input structures are permissible, from the perspective of the skill processing component(s)125, to be used to invoke the skill. The alternate user input structures may be derived by one or more trained models during model training and/or may be based on user input structures provided by different skills. The skill processing component(s)125 associated with a particular skill may also provide the system(s)120 with training text data indicating grammar and annotations. The system(s)120 may use the training text data representing the sample user inputs, the determined related user input(s), the grammar, and the annotations to train a model(s) that indicates when a user input is likely to be directed to/handled by a skill, based at least in part on the structure of the user input. Each trained model of theshortlister component750 may be trained with respect to a different skill. Alternatively, theshortlister component750 may use one trained model per domain, such as one trained model for skills associated with a weather domain, one trained model for skills associated with a ride sharing domain, etc.

The system(s)120 may use the sample user inputs provided by a skill processing component(s)125, and related sample user inputs potentially determined during training, as binary examples to train a model associated with a skill associated with the skill processing component(s)125. The model associated with the particular skill may then be operated at runtime by theshortlister component750. For example, some sample user inputs may be positive examples (e.g., user inputs that may be used to invoke the skill). Other sample user inputs may be negative examples (e.g., user inputs that may not be used to invoke the skill).

As described above, theshortlister component750 may include a different trained model for each skill of the system, a different trained model for each domain, or some other combination of trained model(s). For example, theshortlister component750 may alternatively include a single model. The single model may include a portion trained with respect to characteristics (e.g., semantic characteristics) shared by all skills of the system. The single model may also include skill-specific portions, with each skill-specific portion being trained with respect to a specific skill of the system. Implementing a single model with skill-specific portions may result in less latency than implementing a different trained model for each skill because the single model with skill-specific portions limits the number of characteristics processed on a per skill level.

The portion trained with respect to characteristics shared by more than one skill may be clustered based on domain. For example, a first portion of the portion trained with respect to multiple skills may be trained with respect to weather domain skills, a second portion of the portion trained with respect to multiple skills may be trained with respect to music domain skills, a third portion of the portion trained with respect to multiple skills may be trained with respect to travel domain skills, etc.

Clustering may not be beneficial in every instance because it may cause theshortlister component750 to output indications of only a portion of the skills that theASR output data615 may relate to. For example, a user input may correspond to “tell me about Tom Collins.” If the model is clustered based on domain, theshortlister component750 may determine the user input corresponds to a recipe skill (e.g., a drink recipe) even though the user input may also correspond to an information skill (e.g., including information about a person named Tom Collins).

TheNLU component460 may include one ormore recognizers763. In at least some embodiments, arecognizer763 may be associated with a skill processing component(s)125 (e.g., the recognizer may be configured to interpret text data to correspond to the skill processing component(s)125). In at least some other examples, arecognizer763 may be associated with a domain such as smart home, video, music, weather, custom, etc. (e.g., the recognizer may be configured to interpret text data to correspond to the domain).

If theshortlister component750 determinesASR output data615 is potentially associated with multiple domains, therecognizers763 associated with the domains may process theASR output data615, whilerecognizers763 not indicated in theshortlister component750's output may not process theASR output data615. The “shortlisted”recognizers763 may process theASR output data615 in parallel, in series, partially in parallel, etc. For example, ifASR output data615 potentially relates to both a communications domain and a music domain, a recognizer associated with the communications domain may process theASR output data615 in parallel, or partially in parallel, with a recognizer associated with the music domain processing theASR output data615.

Eachrecognizer763 may include a named entity recognition (NER)component762. TheNER component762 attempts to identify grammars and lexical information that may be used to construe meaning with respect to text data input therein. TheNER component762 identifies portions of text data that correspond to a named entity associated with a domain, associated with therecognizer763 implementing theNER component762. The NER component762 (or other component of the NLU component460) may also determine whether a word refers to an entity whose identity is not explicitly mentioned in the text data, for example “him,” “her,” “it” or other anaphora, exophora, or the like.

Eachrecognizer763, and more specifically eachNER component762, may be associated with a particular grammar database776 and a particular set of intents/actions774 that may be stored in an NLU storage773, and a particular personalized lexicon786 that may be stored in anentity library782. Each gazetteer784 may include domain/skill-indexed lexical information associated with a particular user and/ordevice110. For example, a Gazetteer A (784a) includes skill-indexed lexical information786aato786an. A user's music domain lexical information might include album titles, artist names, and song names, for example, whereas a user's communications domain lexical information might include the names of contacts. Since every user's music collection and contact list is presumably different. This personalized information improves later performed entity resolution.

AnNER component762 applies grammar information776 and lexical information786 associated with a domain (associated with therecognizer763 implementing the NER component762) to determine a mention of one or more entities in text data. In this manner, theNER component762 identifies “slots” (each corresponding to one or more particular words in text data) that may be useful for later processing. TheNER component762 may also label each slot with a type (e.g., noun, place, city, artist name, song name, etc.).

Each grammar database776 includes the names of entities (i.e., nouns) commonly found in speech about the particular domain to which the grammar database776 relates, whereas the lexical information786 is personalized to the user and/or thedevice110 from which the user input originated. For example, a grammar database776 associated with a shopping domain may include a database of words commonly used when people discuss shopping.

A downstream process called entity resolution (discussed in detail elsewhere herein) links a slot of text data to a specific entity known to the system. To perform entity resolution, theNLU component460 may utilize gazetteer information (784a-784n) stored in anentity library storage782. The gazetteer information784 may be used to match text data (representing a portion of the user input) with text data representing known entities, such as song titles, contact names, etc. Gazetteers784 may be linked to users (e.g., a particular gazetteer may be associated with a specific user's music collection), may be linked to certain domains (e.g., a shopping domain, a music domain, a video domain, etc.), or may be organized in a variety of other ways.

Eachrecognizer763 may also include an intent classification (IC)component764. AnIC component764 parses text data to determine an intent(s) (associated with the domain associated with therecognizer763 implementing the IC component764) that potentially represents the user input. An intent represents to an action a user desires be performed. AnIC component764 may communicate with a database774 of words linked to intents. For example, a music intent database may link words and phrases such as “quiet,” “volume off,” and “mute” to a <Mute> intent. AnIC component764 identifies potential intents by comparing words and phrases in text data (representing at least a portion of the user input) to the words and phrases in an intents database774 (associated with the domain that is associated with therecognizer763 implementing the IC component764).

The intents identifiable by aspecific IC component764 are linked to domain-specific (i.e., the domain associated with therecognizer763 implementing the IC component764) grammar frameworks776 with “slots” to be filled. Each slot of a grammar framework776 corresponds to a portion of text data that the system believes corresponds to an entity. For example, a grammar framework776 corresponding to a <PlayMusic> intent may correspond to text data sentence structures such as “Play {Artist Name},” “Play {Album Name},” “Play {Song name},” “Play {Song name} by {Artist Name},” etc. However, to make entity resolution more flexible, grammar frameworks776 may not be structured as sentences, but rather based on associating slots with grammatical tags.

For example, anNER component762 may parse text data to identify words as subject, object, verb, preposition, etc. based on grammar rules and/or models prior to recognizing named entities in the text data. An IC component764 (implemented by thesame recognizer763 as the NER component762) may use the identified verb to identify an intent. TheNER component762 may then determine a grammar model776 associated with the identified intent. For example, a grammar model776 for an intent corresponding to <PlayMusic> may specify a list of slots applicable to play the identified “object” and any object modifier (e.g., a prepositional phrase), such as {Artist Name}, {Album Name}, {Song name}, etc. TheNER component762 may then search corresponding fields in a lexicon786 (associated with the domain associated with therecognizer763 implementing the NER component762), attempting to match words and phrases in text data theNER component762 previously tagged as a grammatical object or object modifier with those identified in the lexicon786.

AnNER component762 may perform semantic tagging, which is the labeling of a word or combination of words according to their type/semantic meaning. AnNER component762 may parse text data using heuristic grammar rules, or a model may be constructed using techniques such as Hidden Markov Models, maximum entropy models, log linear models, conditional random fields (CRF), and the like. For example, anNER component762 implemented by a music domain recognizer may parse and tag text data corresponding to “play mother's little helper by the rolling stones” as {Verb}: “Play,” {Object}: “mother's little helper,” {Object Preposition}: “by,” and {Object Modifier}: “the rolling stones.” TheNER component762 identifies “Play” as a verb based on a word database associated with the music domain, which an IC component764 (also implemented by the music domain recognizer) may determine corresponds to a <PlayMusic> intent. At this stage, no determination has been made as to the meaning of “mother's little helper” or “the rolling stones,” but based on grammar rules and models, theNER component762 has determined the text of these phrases relates to the grammatical object (i.e., entity) of the user input represented in the text data.

AnNER component762 may tag text data to attribute meaning thereto. For example, anNER component762 may tag “play mother's little helper by the rolling stones” as: {domain} Music, {intent}<PlayMusic>, {artist name} rolling stones, {media type} SONG, and {song title} mother's little helper. For further example, theNER component762 may tag “play songs by the rolling stones” as: {domain} Music, {intent}<PlayMusic>, {artist name} rolling stones, and {media type} SONG.

Theshortlister component750 may receiveASR output data615 output from theASR component450 or output from thedevice110b(as illustrated inFIG.8). TheASR component450 may embed theASR output data615 into a form processable by a trained model(s) using sentence embedding techniques as known in the art. Sentence embedding results in theASR output data615 including text in a structure that enables the trained models of theshortlister component850 to operate on theASR output data615. For example, an embedding of theASR output data615 may be a vector representation of theASR output data615.

Theshortlister component750 may make binary determinations (e.g., yes or no) regarding which domains relate to theASR output data615. Theshortlister component750 may make such determinations using the one or more trained models described herein above. If theshortlister component750 implements a single trained model for each domain, theshortlister component750 may simply run the models that are associated with enabled domains as indicated in a user profile associated with thedevice110 and/or user that originated the user input.

Theshortlister component750 may generate n-best list data815 representing domains that may execute with respect to the user input represented in theASR output data615. The size of the n-best list represented in the n-best list data815 is configurable. In an example, the n-best list data815 may indicate every domain of the system as well as contain an indication, for each domain, regarding whether the domain is likely capable to execute the user input represented in theASR output data615. In another example, instead of indicating every domain of the system, the n-best list data815 may only indicate the domains that are likely to be able to execute the user input represented in theASR output data615. In yet another example, theshortlister component750 may implement thresholding such that the n-best list data815 may indicate no more than a maximum number of domains that may execute the user input represented in theASR output data615. In an example, the threshold number of domains that may be represented in the n-best list data815 is ten. In another example, the domains included in the n-best list data815 may be limited by a threshold a score, where only domains indicating a likelihood to handle the user input is above a certain score (as determined by processing theASR output data615 by theshortlister component750 relative to such domains) are included in the n-best list data815.

TheASR output data615 may correspond to more than one ASR hypothesis. When this occurs, theshortlister component750 may output a different n-best list (represented in the n-best list data815) for each ASR hypothesis. Alternatively, theshortlister component750 may output a single n-best list representing the domains that are related to the multiple ASR hypotheses represented in theASR output data615.

As indicated above, theshortlister component750 may implement thresholding such that an n-best list output therefrom may include no more than a threshold number of entries. If theASR output data615 includes more than one ASR hypothesis, the n-best list output by theshortlister component750 may include no more than a threshold number of entries irrespective of the number of ASR hypotheses output by theASR component450. Alternatively or in addition, the n-best list output by theshortlister component750 may include no more than a threshold number of entries for each ASR hypothesis (e.g., no more than five entries for a first ASR hypothesis, no more than five entries for a second ASR hypothesis, etc.).

In addition to making a binary determination regarding whether a domain potentially relates to theASR output data615, theshortlister component750 may generate confidence scores representing likelihoods that domains relate to theASR output data615. If theshortlister component750 implements a different trained model for each domain, theshortlister component750 may generate a different confidence score for each individual domain trained model that is run. If theshortlister component750 runs the models of every domain whenASR output data615 is received, theshortlister component750 may generate a different confidence score for each domain of the system. If theshortlister component750 runs the models of only the domains that are associated with skills indicated as enabled in a user profile associated with thedevice110 and/or user that originated the user input, theshortlister component750 may only generate a different confidence score for each domain associated with at least one enabled skill. If theshortlister component750 implements a single trained model with domain specifically trained portions, theshortlister component750 may generate a different confidence score for each domain who's specifically trained portion is run. Theshortlister component750 may perform matrix vector modification to obtain confidence scores for all domains of the system in a single instance of processing of theASR output data615.

N-best list data815 including confidence scores that may be output by theshortlister component750 may be represented as, for example:

• • Search domain, 0.67
  • Recipe domain, 0.62
  • Information domain, 0.57
  • Shopping domain, 0.42
    As indicated, the confidence scores output by theshortlister component750 may be numeric values. The confidence scores output by theshortlister component750 may alternatively be binned values (e.g., high, medium, low).

The n-best list may only include entries for domains having a confidence score satisfying (e.g., equaling or exceeding) a minimum threshold confidence score. Alternatively, theshortlister component750 may include entries for all domains associated with user enabled skills, even if one or more of the domains are associated with confidence scores that do not satisfy the minimum threshold confidence score.

Theshortlister component750 may considerother data820 when determining which domains may relate to the user input represented in theASR output data615 as well as respective confidence scores. Theother data820 may include usage history data associated with thedevice110 and/or user that originated the user input. For example, a confidence score of a domain may be increased if user inputs originated by thedevice110 and/or user routinely invoke the domain. Conversely, a confidence score of a domain may be decreased if user inputs originated by thedevice110 and/or user rarely invoke the domain. Thus, theother data820 may include an indicator of the user associated with theASR output data615, for example as determined by a user recognition component.

Theother data820 may be character embedded prior to being input to theshortlister component750. Theother data820 may alternatively be embedded using other techniques known in the art prior to being input to theshortlister component750.

Theother data820 may also include data indicating the domains associated with skills that are enabled with respect to thedevice110 and/or user that originated the user input. Theshortlister component750 may use such data to determine which domain-specific trained models to run. That is, theshortlister component750 may determine to only run the trained models associated with domains that are associated with user-enabled skills. Theshortlister component750 may alternatively use such data to alter confidence scores of domains.

As an example, considering two domains, a first domain associated with at least one enabled skill and a second domain not associated with any user-enabled skills of the user that originated the user input, theshortlister component750 may run a first model specific to the first domain as well as a second model specific to the second domain. Alternatively, theshortlister component750 may run a model configured to determine a score for each of the first and second domains. Theshortlister component750 may determine a same confidence score for each of the first and second domains in the first instance. Theshortlister component750 may then alter those confidence scores based on which domains is associated with at least one skill enabled by the present user. For example, theshortlister component750 may increase the confidence score associated with the domain associated with at least one enabled skill while leaving the confidence score associated with the other domain the same. Alternatively, theshortlister component750 may leave the confidence score associated with the domain associated with at least one enabled skill the same while decreasing the confidence score associated with the other domain. Moreover, theshortlister component750 may increase the confidence score associated with the domain associated with at least one enabled skill as well as decrease the confidence score associated with the other domain.

As indicated, a user profile may indicate which skills a corresponding user has enabled (e.g., authorized to execute using data associated with the user). Such indications may be stored in theprofile storage470. When theshortlister component750 receives theASR output data615, theshortlister component750 may determine whether profile data associated with the user and/ordevice110 that originated the command includes an indication of enabled skills.

Theother data820 may also include data indicating the type of thedevice110. The type of a device may indicate the output capabilities of the device. For example, a type of device may correspond to a device with a visual display, a headless (e.g., displayless) device, whether a device is mobile or stationary, whether a device includes audio playback capabilities, whether a device includes a camera, other device hardware configurations, etc. Theshortlister component750 may use such data to determine which domain-specific trained models to run. For example, if thedevice110 corresponds to a displayless type device, theshortlister component750 may determine not to run trained models specific to domains that output video data. Theshortlister component750 may alternatively use such data to alter confidence scores of domains.

As an example, considering two domains, one that outputs audio data and another that outputs video data, theshortlister component750 may run a first model specific to the domain that generates audio data as well as a second model specific to the domain that generates video data. Alternatively theshortlister component750 may run a model configured to determine a score for each domain. Theshortlister component750 may determine a same confidence score for each of the domains in the first instance. Theshortlister component750 may then alter the original confidence scores based on the type of thedevice110 that originated the user input corresponding to theASR output data615. For example, if thedevice110 is a displayless device, theshortlister component750 may increase the confidence score associated with the domain that generates audio data while leaving the confidence score associated with the domain that generates video data the same. Alternatively, if thedevice110 is a displayless device, theshortlister component750 may leave the confidence score associated with the domain that generates audio data the same while decreasing the confidence score associated with the domain that generates video data. Moreover, if thedevice110 is a displayless device, theshortlister component750 may increase the confidence score associated with the domain that generates audio data as well as decrease the confidence score associated with the domain that generates video data.

The type of device information represented in theother data820 may represent output capabilities of the device to be used to output content to the user, which may not necessarily be the user input originating device. For example, a user may input a spoken user input corresponding to “play Game of Thrones” to a device not including a display. The system may determine a smart TV or other display device (associated with the same user profile) for outputting Game of Thrones. Thus, theother data820 may represent the smart TV of other display device, and not the displayless device that captured the spoken user input.

Theother data820 may also include data indicating the user input originating device's speed, location, or other mobility information. For example, the device may correspond to a vehicle including a display. If the vehicle is moving, theshortlister component750 may decrease the confidence score associated with a domain that generates video data as it may be undesirable to output video content to a user while the user is driving. The device may output data to the system(s)120 indicating when the device is moving.

Theother data820 may also include data indicating a currently invoked domain. For example, a user may speak a first (e.g., a previous) user input causing the system to invoke a music domain skill to output music to the user. As the system is outputting music to the user, the system may receive a second (e.g., the current) user input. Theshortlister component750 may use such data to alter confidence scores of domains. For example, theshortlister component750 may run a first model specific to a first domain as well as a second model specific to a second domain. Alternatively, theshortlister component750 may run a model configured to determine a score for each domain. Theshortlister component750 may also determine a same confidence score for each of the domains in the first instance. Theshortlister component750 may then alter the original confidence scores based on the first domain being invoked to cause the system to output content while the current user input was received. Based on the first domain being invoked, theshortlister component750 may (i) increase the confidence score associated with the first domain while leaving the confidence score associated with the second domain the same, (ii) leave the confidence score associated with the first domain the same while decreasing the confidence score associated with the second domain, or (iii) increase the confidence score associated with the first domain as well as decrease the confidence score associated with the second domain.

The thresholding implemented with respect to the n-best list data815 generated by theshortlister component750 as well as the different types ofother data820 considered by theshortlister component750 are configurable. For example, theshortlister component750 may update confidence scores as moreother data820 is considered. For further example, the n-best list data815 may exclude relevant domains if thresholding is implemented. Thus, for example, theshortlister component750 may include an indication of a domain in the n-best list815 unless theshortlister component750 is one hundred percent confident that the domain may not execute the user input represented in the ASR output data615 (e.g., theshortlister component750 determines a confidence score of zero for the domain).

Theshortlister component750 may send theASR output data615 torecognizers763 associated with domains represented in the n-best list data815. Alternatively, theshortlister component750 may send the n-best list data815 or some other indicator of the selected subset of domains to another component (such as the orchestrator component430) which may in turn send theASR output data615 to therecognizers763 corresponding to the domains included in the n-best list data815 or otherwise indicated in the indicator. If theshortlister component750 generates an n-best list representing domains without any associated confidence scores, theshortlister component750/orchestrator component430 may send theASR output data615 torecognizers763 associated with domains that theshortlister component750 determines may execute the user input. If theshortlister component750 generates an n-best list representing domains with associated confidence scores, theshortlister component750/orchestrator component430 may send theASR output data615 torecognizers763 associated with domains associated with confidence scores satisfying (e.g., meeting or exceeding) a threshold minimum confidence score.

Arecognizer763 may output tagged text data generated by anNER component762 and anIC component764, as described herein above. TheNLU component460 may compile the output tagged text data of therecognizers763 into a single cross-domain n-best list840 and may send the cross-domain n-best list840 to apruning component850. Each entry of tagged text (e.g., each NLU hypothesis) represented in the cross-domain n-best list data840 may be associated with a respective score indicating a likelihood that the NLU hypothesis corresponds to the domain associated with therecognizer763 from which the NLU hypothesis was output. For example, the cross-domain n-best list data840 may be represented as (with each line corresponding to a different NLU hypothesis):

• • [0.95] Intent: <PlayMusic> ArtistName: Beethoven SongName: Waldstein Sonata
  • [0.70] Intent: <PlayVideo> ArtistName: Beethoven VideoName: Waldstein Sonata
  • [0.01] Intent: <PlayMusic> ArtistName: Beethoven AlbumName: Waldstein Sonata
  • [0.01] Intent: <PlayMusic> SongName: Waldstein Sonata

Thepruning component850 may sort the NLU hypotheses represented in the cross-domain n-best list data840 according to their respective scores. Thepruning component850 may perform score thresholding with respect to the cross-domain NLU hypotheses. For example, thepruning component850 may select NLU hypotheses associated with scores satisfying (e.g., meeting and/or exceeding) a threshold score. Thepruning component850 may also or alternatively perform number of NLU hypothesis thresholding. For example, thepruning component850 may select the top scoring NLU hypothesis(es). Thepruning component850 may output a portion of the NLU hypotheses input thereto. The purpose of thepruning component850 is to create a reduced list of NLU hypotheses so that downstream, more resource intensive, processes may only operate on the NLU hypotheses that most likely represent the user's intent.

TheNLU component460 may include a lightslot filler component852. The lightslot filler component852 can take text from slots represented in the NLU hypotheses output by thepruning component850 and alter them to make the text more easily processed by downstream components. The lightslot filler component852 may perform low latency operations that do not involve heavy operations such as reference to a knowledge base (e.g.,772. The purpose of the lightslot filler component852 is to replace words with other words or values that may be more easily understood by downstream components. For example, if a NLU hypothesis includes the word “tomorrow,” the lightslot filler component852 may replace the word “tomorrow” with an actual date for purposes of downstream processing. Similarly, the lightslot filler component852 may replace the word “CD” with “album” or the words “compact disc.” The replaced words are then included in the cross-domain n-best list data860.

The cross-domain n-best list data860 may be input to anentity resolution component870. Theentity resolution component870 can apply rules or other instructions to standardize labels or tokens from previous stages into an intent/slot representation. The precise transformation may depend on the domain. For example, for a travel domain, theentity resolution component870 may transform text corresponding to “Boston airport” to the standard BOS three-letter code referring to the airport. Theentity resolution component870 can refer to a knowledge base (e.g.,772) that is used to specifically identify the precise entity referred to in each slot of each NLU hypothesis represented in the cross-domain n-best list data860. Specific intent/slot combinations may also be tied to a particular source, which may then be used to resolve the text. In the example “play songs by the stones,” theentity resolution component870 may reference a personal music catalog, Amazon Music account, a user profile, or the like. Theentity resolution component870 may output an altered n-best list that is based on the cross-domain n-best list860 but that includes more detailed information (e.g., entity IDs) about the specific entities mentioned in the slots and/or more detailed slot data that can eventually be used by a skill. TheNLU component460 may include multipleentity resolution components870 and eachentity resolution component870 may be specific to one or more domains.

TheNLU component460 may include areranker890. Thereranker890 may assign a particular confidence score to each NLU hypothesis input therein. The confidence score of a particular NLU hypothesis may be affected by whether the NLU hypothesis has unfilled slots. For example, if a NLU hypothesis includes slots that are all filled/resolved, that NLU hypothesis may be assigned a higher confidence score than another NLU hypothesis including at least some slots that are unfilled/unresolved by theentity resolution component870.

Thereranker890 may apply re-scoring, biasing, or other techniques. Thereranker890 may consider not only the data output by theentity resolution component870, but may also considerother data891. Theother data891 may include a variety of information. For example, theother data891 may include skill rating or popularity data. For example, if one skill has a high rating, thereranker890 may increase the score of a NLU hypothesis that may be processed by the skill. Theother data891 may also include information about skills that have been enabled by the user that originated the user input. For example, thereranker890 may assign higher scores to NLU hypothesis that may be processed by enabled skills than NLU hypothesis that may be processed by non-enabled skills. Theother data891 may also include data indicating user usage history, such as if the user that originated the user input regularly uses a particular skill or does so at particular times of day. Theother data891 may additionally include data indicating date, time, location, weather, type ofdevice110, user identifier, context, as well as other information. For example, thereranker890 may consider when any particular skill is currently active (e.g., music being played, a game being played, etc.).

As illustrated and described, theentity resolution component870 is implemented prior to thereranker890. Theentity resolution component870 may alternatively be implemented after thereranker890. Implementing theentity resolution component870 after thereranker890 limits the NLU hypotheses processed by theentity resolution component870 to only those hypotheses that successfully pass through thereranker890.

Thereranker890 may be a global reranker (e.g., one that is not specific to any particular domain). Alternatively, theNLU component460 may implement one or more domain-specific rerankers. Each domain-specific reranker may rerank NLU hypotheses associated with the domain. Each domain-specific reranker may output an n-best list of reranked hypotheses (e.g., 5-10 hypotheses).

TheNLU component460 may perform NLU processing described above with respect to domains associated with skills wholly implemented as part of the system(s)120 (e.g., designated490 inFIG.4). TheNLU component460 may separately perform NLU processing described above with respect to domains associated with skills that are at least partially implemented as part of the skill processing component(s)125. In an example, theshortlister component750 may only process with respect to these latter domains. Results of these two NLU processing paths may be merged intoNLU output data885, which may be sent to apost-NLU ranker465, which may be implemented by the system(s)120.

Thepost-NLU ranker465 may include a statistical component that produces a ranked list of intent/skill pairs with associated confidence scores. Each confidence score may indicate an adequacy of the skill's execution of the intent with respect to NLU results data associated with the skill. Thepost-NLU ranker465 may operate one or more trained models configured to process theNLU results data885,skill result data830, and theother data820 in order to output rankedoutput data825. The rankedoutput data825 may include an n-best list where the NLU hypotheses in theNLU results data885 are reordered such that the n-best list in the rankedoutput data825 represents a prioritized list of skills to respond to a user input as determined by thepost-NLU ranker465. The rankedoutput data825 may also include (either as part of an n-best list or otherwise) individual respective scores corresponding to skills where each score indicates a probability that the skill (and/or its respective result data) corresponds to the user input.

The system may be configured with thousands, tens of thousands, etc. skills. Thepost-NLU ranker465 enables the system to better determine the best skill to execute the user input. For example, first and second NLU hypotheses in theNLU results data885 may substantially correspond to each other (e.g., their scores may be significantly similar), even though the first NLU hypothesis may be processed by a first skill and the second NLU hypothesis may be processed by a second skill. The first NLU hypothesis may be associated with a first confidence score indicating the system's confidence with respect to NLU processing performed to generate the first NLU hypothesis. Moreover, the second NLU hypothesis may be associated with a second confidence score indicating the system's confidence with respect to NLU processing performed to generate the second NLU hypothesis. The first confidence score may be similar or identical to the second confidence score. The first confidence score and/or the second confidence score may be a numeric value (e.g., from 0.0 to 1.0). Alternatively, the first confidence score and/or the second confidence score may be a binned value (e.g., low, medium, high).

The post-NLU ranker465 (or other scheduling component such as orchestrator component430) may solicit the first skill and the second skill to providepotential result data830 based on the first NLU hypothesis and the second NLU hypothesis, respectively. For example, thepost-NLU ranker465 may send the first NLU hypothesis to the first skill490aalong with a request for the first skill490ato at least partially execute with respect to the first NLU hypothesis. Thepost-NLU ranker465 may also send the second NLU hypothesis to the second skill490balong with a request for the second skill490bto at least partially execute with respect to the second NLU hypothesis. Thepost-NLU ranker465 receives, from the first skill490a, first result data830agenerated from the first skill490a's execution with respect to the first NLU hypothesis. Thepost-NLU ranker465 also receives, from the second skill490b, second results data830bgenerated from the second skill490b's execution with respect to the second NLU hypothesis.

Theresult data830 may include various portions. For example, theresult data830 may include content (e.g., audio data, text data, and/or video data) to be output to a user. Theresult data830 may also include a unique identifier used by the system(s)120 and/or the skill processing component(s)125 to locate the data to be output to a user. Theresult data830 may also include an instruction. For example, if the user input corresponds to “turn on the light,” theresult data830 may include an instruction causing the system to turn on a light associated with a profile of the device (110a/110b) and/or user.

Thepost-NLU ranker465 may consider the first result data830aand the second result data830bto alter the first confidence score and the second confidence score of the first NLU hypothesis and the second NLU hypothesis, respectively. That is, thepost-NLU ranker465 may generate a third confidence score based on the first result data830aand the first confidence score. The third confidence score may correspond to how likely thepost-NLU ranker465 determines the first skill will correctly respond to the user input. Thepost-NLU ranker465 may also generate a fourth confidence score based on the second result data830band the second confidence score. One skilled in the art will appreciate that a first difference between the third confidence score and the fourth confidence score may be greater than a second difference between the first confidence score and the second confidence score. Thepost-NLU ranker465 may also consider theother data820 to generate the third confidence score and the fourth confidence score. While it has been described that thepost-NLU ranker465 may alter the confidence scores associated with first and second NLU hypotheses, one skilled in the art will appreciate that thepost-NLU ranker465 may alter the confidence scores of more than two NLU hypotheses. Thepost-NLU ranker465 may select theresult data830 associated with the skill490 with the highest altered confidence score to be the data output in response to the current user input. Thepost-NLU ranker465 may also consider theASR output data615 to alter the NLU hypotheses confidence scores.

Theorchestrator component430 may, prior to sending theNLU results data885 to thepost-NLU ranker465, associate intents in the NLU hypotheses with skills490. For example, if a NLU hypothesis includes a <PlayMusic> intent, theorchestrator component430 may associate the NLU hypothesis with one or more skills490 that can execute the <PlayMusic> intent. Thus, theorchestrator component430 may send theNLU results data885, including NLU hypotheses paired with skills490, to thepost-NLU ranker465. In response toASR output data615 corresponding to “what should I do for dinner today,” theorchestrator component430 may generates pairs of skills490 with associated NLU hypotheses corresponding to:

• • Skill 1/NLU hypothesis including <Help> intent
  • Skill 2/NLU hypothesis including <Order> intent
  • Skill 3/NLU hypothesis including <DishType> intent

Thepost-NLU ranker465 queries each skill490, paired with a NLU hypothesis in theNLU output data885, to provideresult data830 based on the NLU hypothesis with which it is associated. That is, with respect to each skill, thepost-NLU ranker465 colloquially asks the each skill “if given this NLU hypothesis, what would you do with it.” According to the above example, thepost-NLU ranker465 may send skills490 the following data:

• • Skill 1: First NLU hypothesis including <Help> intent indicator
  • Skill 2: Second NLU hypothesis including <Order> intent indicator
  • Skill 3: Third NLU hypothesis including <DishType> intent indicator Thepost-NLU ranker465 may query each of the skills490 in parallel or substantially in parallel.

A skill490 may provide thepost-NLU ranker465 with various data and indications in response to thepost-NLU ranker465 soliciting the skill490 forresult data830. A skill490 may simply provide thepost-NLU ranker465 with an indication of whether or not the skill can execute with respect to the NLU hypothesis it received. A skill490 may also or alternatively provide thepost-NLU ranker465 with output data generated based on the NLU hypothesis it received. In some situations, a skill490 may need further information in addition to what is represented in the received NLU hypothesis to provide output data responsive to the user input. In these situations, the skill490 may provide thepost-NLU ranker465 withresult data830 indicating slots of a framework that the skill490 further needs filled or entities that the skill490 further needs resolved prior to the skill490 being able to providedresult data830 responsive to the user input. The skill490 may also provide thepost-NLU ranker465 with an instruction and/or computer-generated speech indicating how the skill490 recommends the system solicit further information needed by the skill490. The skill490 may further provide thepost-NLU ranker465 with an indication of whether the skill490 will have all needed information after the user provides additional information a single time, or whether the skill490 will need the user to provide various kinds of additional information prior to the skill490 having all needed information. According to the above example, skills490 may provide thepost-NLU ranker465 with the following:

• • Skill 1: indication representing the skill can execute with respect to a NLU hypothesis including the <Help> intent indicator
  • Skill 2: indication representing the skill needs to the system to obtain further information
  • Skill 3: indication representing the skill can provide numerous results in response to the third NLU hypothesis including the <DishType> intent indicator

Result data830 includes an indication provided by a skill490 indicating whether or not the skill490 can execute with respect to a NLU hypothesis; data generated by a skill490 based on a NLU hypothesis; as well as an indication provided by a skill490 indicating the skill490 needs further information in addition to what is represented in the received NLU hypothesis.

Thepost-NLU ranker465 uses theresult data830 provided by the skills490 to alter the NLU processing confidence scores generated by thereranker890. That is, thepost-NLU ranker465 uses theresult data830 provided by the queried skills490 to create larger differences between the NLU processing confidence scores generated by thereranker890. Without thepost-NLU ranker465, the system may not be confident enough to determine an output in response to a user input, for example when the NLU hypotheses associated with multiple skills are too close for the system to confidently determine a single skill490 to invoke to respond to the user input. For example, if the system does not implement thepost-NLU ranker465, the system may not be able to determine whether to obtain output data from a general reference information skill or a medical information skill in response to a user input corresponding to “what is acne.”

Thepost-NLU ranker465 may prefer skills490 that provideresult data830 responsive to NLU hypotheses over skills490 that provideresult data830 corresponding to an indication that further information is needed, as well as skills490 that provideresult data830 indicating they can provide multiple responses to received NLU hypotheses. For example, thepost-NLU ranker465 may generate a first score for a first skill490athat is greater than the first skill's NLU confidence score based on the first skill490aproviding result data830aincluding a response to a NLU hypothesis. For further example, thepost-NLU ranker465 may generate a second score for a second skill490bthat is less than the second skill's NLU confidence score based on the second skill490bproviding result data830bindicating further information is needed for the second skill490bto provide a response to a NLU hypothesis. Yet further, for example, thepost-NLU ranker465 may generate a third score for a third skill490cthat is less than the third skill's NLU confidence score based on the third skill490cproviding result data830cindicating the third skill490ccan provide multiple responses to a NLU hypothesis.

Thepost-NLU ranker465 may considerother data820 in determining scores. Theother data820 may include rankings associated with the queried skills490. A ranking may be a system ranking or a user-specific ranking. A ranking may indicate a veracity of a skill from the perspective of one or more users of the system. For example, thepost-NLU ranker465 may generate a first score for a first skill490athat is greater than the first skill's NLU processing confidence score based on the first skill490abeing associated with a high ranking. For further example, thepost-NLU ranker465 may generate a second score for a second skill490bthat is less than the second skill's NLU processing confidence score based on the second skill490bbeing associated with a low ranking.

Theother data820 may include information indicating whether or not the user that originated the user input has enabled one or more of the queried skills490. For example, thepost-NLU ranker465 may generate a first score for a first skill490athat is greater than the first skill's NLU processing confidence score based on the first skill490abeing enabled by the user that originated the user input. For further example, thepost-NLU ranker465 may generate a second score for a second skill490bthat is less than the second skill's NLU processing confidence score based on the second skill490bnot being enabled by the user that originated the user input. When thepost-NLU ranker465 receives theNLU results data885, thepost-NLU ranker465 may determine whether profile data, associated with the user and/or device that originated the user input, includes indications of enabled skills.

Theother data820 may include information indicating output capabilities of a device that will be used to output content, responsive to the user input, to the user. The system may include devices that include speakers but not displays, devices that include displays but not speakers, and devices that include speakers and displays. If the device that will output content responsive to the user input includes one or more speakers but not a display, thepost-NLU ranker465 may increase the NLU processing confidence score associated with a first skill configured to output audio data and/or decrease the NLU processing confidence score associated with a second skill configured to output visual data (e.g., image data and/or video data). If the device that will output content responsive to the user input includes a display but not one or more speakers, thepost-NLU ranker465 may increase the NLU processing confidence score associated with a first skill configured to output visual data and/or decrease the NLU processing confidence score associated with a second skill configured to output audio data.

Theother data820 may include information indicating the veracity of theresult data830 provided by a skill490. For example, if a user says “tell me a recipe for pasta sauce,” a first skill490amay provide thepost-NLU ranker465 with first result data830acorresponding to a first recipe associated with a five star rating and a second skill490bmay provide thepost-NLU ranker465 with second result data830bcorresponding to a second recipe associated with a one star rating. In this situation, thepost-NLU ranker465 may increase the NLU processing confidence score associated with the first skill490abased on the first skill490aproviding the first result data830aassociated with the five star rating and/or decrease the NLU processing confidence score associated with the second skill490bbased on the second skill490bproviding the second result data830bassociated with the one star rating.

Theother data820 may include information indicating the type of device that originated the user input. For example, the device may correspond to a “hotel room” type if the device is located in a hotel room. If a user inputs a command corresponding to “order me food” to the device located in the hotel room, thepost-NLU ranker465 may increase the NLU processing confidence score associated with a first skill490acorresponding to a room service skill associated with the hotel and/or decrease the NLU processing confidence score associated with a second skill490bcorresponding to a food skill not associated with the hotel.

Theother data820 may include information indicating a location of the device and/or user that originated the user input. The system may be configured with skills490 that may only operate with respect to certain geographic locations. For example, a user may provide a user input corresponding to “when is the next train to Portland.” A first skill490amay operate with respect to trains that arrive at, depart from, and pass through Portland, Oregon. A second skill490bmay operate with respect to trains that arrive at, depart from, and pass through Portland, Maine. If the device and/or user that originated the user input is located in Seattle, Washington, thepost-NLU ranker465 may increase the NLU processing confidence score associated with the first skill490aand/or decrease the NLU processing confidence score associated with the second skill490b. Likewise, if the device and/or user that originated the user input is located in Boston, Massachusetts, thepost-NLU ranker465 may increase the NLU processing confidence score associated with the second skill490band/or decrease the NLU processing confidence score associated with the first skill490a.

Theother data820 may include information indicating a time of day. The system may be configured with skills490 that operate with respect to certain times of day. For example, a user may provide a user input corresponding to “order me food.” A first skill490amay generate first result data830acorresponding to breakfast. A second skill490bmay generate second result data830bcorresponding to dinner. If the system(s)120 receives the user input in the morning, thepost-NLU ranker465 may increase the NLU processing confidence score associated with the first skill490aand/or decrease the NLU processing score associated with the second skill490b. If the system(s)120 receives the user input in the afternoon or evening, thepost-NLU ranker465 may increase the NLU processing confidence score associated with the second skill490band/or decrease the NLU processing confidence score associated with the first skill490a.

Theother data820 may include information indicating user preferences. The system may include multiple skills490 configured to execute in substantially the same manner. For example, a first skill490aand a second skill490bmay both be configured to order food from respective restaurants. The system may store a user preference (e.g., in the profile storage470) that is associated with the user that provided the user input to the system(s)120 as well as indicates the user prefers the first skill490aover the second skill490b. Thus, when the user provides a user input that may be executed by both the first skill490aand the second skill490b, thepost-NLU ranker465 may increase the NLU processing confidence score associated with the first skill490aand/or decrease the NLU processing confidence score associated with the second skill490b.

Theother data820 may include information indicating system usage history associated with the user that originated the user input. For example, the system usage history may indicate the user originates user inputs that invoke a first skill490amore often than the user originates user inputs that invoke a second skill490b. Based on this, if the present user input may be executed by both the first skill490aand the second skill490b, thepost-NLU ranker465 may increase the NLU processing confidence score associated with the first skill490aand/or decrease the NLU processing confidence score associated with the second skill490b.

Theother data820 may include information indicating a speed at which thedevice110 that originated the user input is traveling. For example, thedevice110 may be located in a moving vehicle, or may be a moving vehicle. When adevice110 is in motion, the system may prefer audio outputs rather than visual outputs to decrease the likelihood of distracting the user (e.g., a driver of a vehicle). Thus, for example, if thedevice110 that originated the user input is moving at or above a threshold speed (e.g., a speed above an average user's walking speed), thepost-NLU ranker465 may increase the NLU processing confidence score associated with a first skill490athat generates audio data. Thepost-NLU ranker465 may also or alternatively decrease the NLU processing confidence score associated with a second skill490bthat generates image data or video data.

Theother data820 may include information indicating how long it took a skill490 to provideresult data830 to thepost-NLU ranker465. When thepost-NLU ranker465 multiple skills490 forresult data830, the skills490 may respond to the queries at different speeds. Thepost-NLU ranker465 may implement a latency budget. For example, if thepost-NLU ranker465 determines a skill490 responds to thepost-NLU ranker465 within a threshold amount of time from receiving a query from thepost-NLU ranker465, thepost-NLU ranker465 may increase the NLU processing confidence score associated with the skill490. Conversely, if thepost-NLU ranker465 determines a skill490 does not respond to thepost-NLU ranker465 within a threshold amount of time from receiving a query from thepost-NLU ranker465, thepost-NLU ranker465 may decrease the NLU processing confidence score associated with the skill490.

It has been described that thepost-NLU ranker465 uses theother data820 to increase and decrease NLU processing confidence scores associated with various skills490 that thepost-NLU ranker465 has already requested result data from. Alternatively, thepost-NLU ranker465 may use theother data820 to determine which skills490 to request result data from. For example, thepost-NLU ranker465 may use theother data820 to increase and/or decrease NLU processing confidence scores associated with skills490 associated with theNLU results data885 output by theNLU component460. Thepost-NLU ranker465 may select n-number of top scoring altered NLU processing confidence scores. Thepost-NLU ranker465 may then requestresult data830 from only the skills490 associated with the selected n-number of NLU processing confidence scores.

As described, thepost-NLU ranker465 may request resultdata830 from all skills490 associated with theNLU results data885 output by theNLU component460. Alternatively, the system(s)120 may prefer resultdata830 from skills implemented entirely by the system(s)120 rather than skills at least partially implemented by the skill processing component(s)125. Therefore, in the first instance, thepost-NLU ranker465 may request resultdata830 from only skills associated with theNLU results data885 and entirely implemented by the system(s)120. Thepost-NLU ranker465 may only requestresult data830 from skills associated with theNLU results data885, and at least partially implemented by the skill processing component(s)125, if none of the skills, wholly implemented by the system(s)120, provide thepost-NLU ranker465 withresult data830 indicating either data response to theNLU results data885, an indication that the skill can execute the user input, or an indication that further information is needed.

As indicated above, thepost-NLU ranker465 may request resultdata830 from multiple skills490. If one of the skills490 providesresult data830 indicating a response to a NLU hypothesis and the other skills provideresult data830 indicating either they cannot execute or they need further information, thepost-NLU ranker465 may select theresult data830 including the response to the NLU hypothesis as the data to be output to the user. If more than one of the skills490 providesresult data830 indicating responses to NLU hypotheses, thepost-NLU ranker465 may consider theother data820 to generate altered NLU processing confidence scores, and select theresult data830 of the skill associated with the greatest score as the data to be output to the user.

A system that does not implement thepost-NLU ranker465 may select the highest scored NLU hypothesis in the NLU resultsdata885. The system may send the NLU hypothesis to a skill490 associated therewith along with a request for output data. In some situations, the skill490 may not be able to provide the system with output data. This results in the system indicating to the user that the user input could not be processed even though another skill associated with lower ranked NLU hypothesis could have provided output data responsive to the user input.

Thepost-NLU ranker465 reduces instances of the aforementioned situation. As described, thepost-NLU ranker465 queries multiple skills associated with theNLU results data885 to provideresult data830 to thepost-NLU ranker465 prior to thepost-NLU ranker465 ultimately determining the skill490 to be invoked to respond to the user input. Some of the skills490 may provideresult data830 indicating responses to NLU hypotheses while other skills490 may providing resultdata830 indicating the skills cannot provide responsive data. Whereas a system not implementing thepost-NLU ranker465 may select one of the skills490 that could not provide a response, thepost-NLU ranker465 only selects a skill490 that provides thepost-NLU ranker465 with result data corresponding to a response, indicating further information is needed, or indicating multiple responses can be generated.

Thepost-NLU ranker465 may selectresult data830, associated with the skill490 associated with the highest score, for output to the user. Alternatively, thepost-NLU ranker465 may output rankedoutput data825 indicating skills490 and their respective post-NLU ranker rankings. Since thepost-NLU ranker465 receivesresult data830, potentially corresponding to a response to the user input, from the skills490 prior topost-NLU ranker465 selecting one of the skills or outputting the rankedoutput data825, little to no latency occurs from the time skills provideresult data830 and the time the system outputs responds to the user.

If thepost-NLU ranker465 selects result audio data to be output to a user and the system determines content should be output audibly, the post-NLU ranker465 (or another component of the system(s)120) may cause thedevice110aand/or thedevice110bto output audio corresponding to the result audio data. If thepost-NLU ranker465 selects result text data to output to a user and the system determines content should be output visually, the post-NLU ranker465 (or another component of the system(s)120) may cause thedevice110bto display text corresponding to the result text data. If thepost-NLU ranker465 selects result audio data to output to a user and the system determines content should be output visually, the post-NLU ranker465 (or another component of the system(s)120) may send the result audio data to theASR component450. TheASR component450 may generate output text data corresponding to the result audio data. The system(s)120 may then cause thedevice110bto display text corresponding to the output text data. If thepost-NLU ranker465 selects result text data to output to a user and the system determines content should be output audibly, the post-NLU ranker465 (or another component of the system(s)120) may send the result text data to theTTS component480. TheTTS component480 may generate output audio data (corresponding to computer-generated speech) based on the result text data. The system(s)120 may then cause thedevice110aand/or thedevice110bto output audio corresponding to the output audio data.

As described, a skill490 may provideresult data830 either indicating a response to the user input, indicating more information is needed for the skill490 to provide a response to the user input, or indicating the skill490 cannot provide a response to the user input. If the skill490 associated with the highest post-NLU ranker score provides thepost-NLU ranker465 withresult data830 indicating a response to the user input, the post-NLU ranker465 (or another component of the system(s)120, such as the orchestrator component430) may simply cause content corresponding to theresult data830 to be output to the user. For example, thepost-NLU ranker465 may send theresult data830 to theorchestrator component430. Theorchestrator component430 may cause theresult data830 to be sent to the device (110a/110b), which may output audio and/or display text corresponding to theresult data830. Theorchestrator component430 may send theresult data830 to theASR component450 to generate output text data and/or may send theresult data830 to theTTS component480 to generate output audio data, depending on the situation.

The skill490 associated with the highest post-NLU ranker score may provide thepost-NLU ranker465 withresult data830 indicating more information is needed as well as instruction data. The instruction data may indicate how the skill490 recommends the system obtain the needed information. For example, the instruction data may correspond to text data or audio data (i.e., computer-generated speech) corresponding to “please indicate______.” The instruction data may be in a format (e.g., text data or audio data) capable of being output by the device (110a/110b). When this occurs, thepost-NLU ranker465 may simply cause the received instruction data be output by the device (110a/110b). Alternatively, the instruction data may be in a format that is not capable of being output by the device (110a/110b). When this occurs, thepost-NLU ranker465 may cause theASR component450 or theTTS component480 to process the instruction data, depending on the situation, to generate instruction data that may be output by the device (110a/110b). Once the user provides the system with all further information needed by the skill490, the skill490 may provide the system withresult data830 indicating a response to the user input, which may be output by the system as detailed above.

The system may include “informational” skills490 that simply provide the system with information, which the system outputs to the user. The system may also include “transactional” skills490 that require a system instruction to execute the user input. Transactional skills490 include ride sharing skills, flight booking skills, etc. A transactional skill490 may simply provide thepost-NLU ranker465 withresult data830 indicating the transactional skill490 can execute the user input. Thepost-NLU ranker465 may then cause the system to solicit the user for an indication that the system is permitted to cause the transactional skill490 to execute the user input. The user-provided indication may be an audible indication or a tactile indication (e.g., activation of a virtual button or input of text via a virtual keyboard). In response to receiving the user-provided indication, the system may provide the transactional skill490 with data corresponding to the indication. In response, the transactional skill490 may execute the command (e.g., book a flight, book a train ticket, etc.). Thus, while the system may not further engage an informational skill490 after the informational skill490 provides thepost-NLU ranker465 withresult data830, the system may further engage a transactional skill490 after the transactional skill490 provides thepost-NLU ranker465 withresult data830 indicating the transactional skill490 may execute the user input.

In some instances, thepost-NLU ranker465 may generate respective scores for first and second skills that are too close (e.g., are not different by at least a threshold difference) for thepost-NLU ranker465 to make a confident determination regarding which skill should execute the user input. When this occurs, the system may request the user indicate which skill the user prefers to execute the user input. The system may output TTS-generated speech to the user to solicit which skill the user wants to execute the user input.

Theother data820 and/orother data891 may include state data195 that represents a state of particular operations relative to a specific speech-processing system120. For example, afirst system120amay have access to firstsystem state data195athat indicates what interactions a particular user profile/device110 have had with thefirst system120awhile asecond system120bmay have access to secondsystem state data195bthat indicates what interactions a particular user profile/device110 have had with thesecond system120b. As can be appreciated, firstsystem state data195awill be different from secondsystem state data195band thefirst system120awill likely not have access to secondsystem state data195bwhile thesecond system120bwill likely not have access to firstsystem state data195a. (And so on for other systems and their respective state data195.) The respective system's state data195 as part ofother data820 and/orother data891 allows aspecific system120 to interpret and/or rank an incoming request in a manner consistent with previous interactions with thatparticular system120, as such interactions may be represented in the state data195.

Speech processing, such as that described above, may be based on the device/profile state data194. Specifically, theother data820 and/orother data891 may also includestate data194 that represents what device process controls may be executed by a requestingdevice110 and/or a device associated with a user profile of a requesting device (for example as indicated bystate data194m). As discussed above,certain state data194 related to adevice110 and/or user profile may be made available to anassistant system120. Such state data194 (e.g.,194a,194b, etc.) may have some overlap withstate data194mthat is available to a device component, for example amulti-assistant component115. The state data made available to afirst assistant system120amay not include information related to asecond assistant system120b. For example,state data194bavailable to asecond system120bmay not include information identifying operations performed by afirst system120a. This may be due to privacy perception concerns, security concerns, system configurations, etc. However, allowing access to certainlimited state data194 enables thesystem100 to allow control of device processes by multiple assistant system(s)120, even those that may not have initiated a particular device process.

As noted above, a device process may involve controlling a process that involves some action to be performed by thedevice110. Such a device process control may include, for example, starting/stopping a timer, setting/stopping an alarm, playing/stopping media content (such as a song, video, podcast, etc.), controlling output content (such as skipping a song, going back a song, extending/snoozing a timer/alarm, stopping synthesized speech output, etc.), setting a temperature (for example if a device may operate as a thermostat), activating/deactivating a component of the device (such as a camera, light, etc.), controlling a device setting (such as volume, brightness, sensitivity, etc.), setting/controlling a reminder, initiating/controlling/terminating a call or call request, or the like. A device process control may thus control a device to transition from a first state (e.g., outputting audio, showing something on a display) to a second state (e.g., ceasing output of audio, outputting audio at a different volume, showing something else on the display, removing something from the display, etc.).

The state data194 (e.g.,194m) included in theother data820 and/orother data891 may thus indicate what such device process controls may be executed by adevice110 and/or other device(s) associated with the particular user profile. To determine thisstate data194, asystem120 may receive thestate data194 from a device, for example from amulti-assistant component115 as illustrated above in reference toFIGS.2A and2B. In another example, asystem120 may receive an indicator of thedevice110/user profile of an incoming request and may obtain theappropriate state data194 from another source using the indicator. For example, thesystem120 may communicate with a storage component corresponding to the device/user profile to obtain therelevant state data194.

Thereranker890 and/orpost-NLU ranker465 may use the device/profile state data194 to interpret theASR data615/select a particular NLU hypothesis that interprets the incoming utterance as one that is requesting control of a device process. For example, a user may speak a command such as “Alexa, stop.” Without access to the device/profile state data194, an NLU component460bof asecond system120bmay determine a potential NLU hypothesis of “stop music” but if thestate data195bdoes not indicate active music playback relative to both thesecond system120band the particular requestingdevice110, the interpretation of “stop music” may receive a low ranking (for example by a reranker890band/or post-NLU ranker465b) because thesecond system120bis unaware of any active music playback. If, however, thesecond system120bhas access to the device/profile state data194, and the device/profile state data194 indicates that the device110 (or another device of the requesting profile) is capable of performing a stop music command, thereranker890 and/orpost-NLU ranker465 may give the interpretation of “stop music” a higher ranking. Thus, in a situation where a “play music” command was previously spoken to an assistant/system other thansecond system120b, thesecond system120bmay still be able to properly interpret a command such as “Alexa, stop” as a command to control a device process controllable by device110 (as indicated by the device/profile state data194) even if thesecond system120bhas no information about the specific active music playback, what music is playing, how it was initiated, etc.

In the above example, with the use of device/profile state data194, the NLU460band/or post-NLU ranker465bmay properly interpret “Alexa, stop” as a command to control a device process. Thus the selected hypothesis may correspond toNLU results data825/885 that indicates a <stopmusic> command and also indicates the destination for such a command should be adevice skill191b. The NLU resultsdata825/885, along with an indicator linked to the requestingdevice110/user profile, may be output by the NLU460band/or post-NLU ranker465b. The indicator may be an indicator of the requestingdevice110, the requesting user profile, the specific utterance, or other indicator that may be used by another component (for example Orchestrator230) to link theNLU results data825/885 to the original request/utterance.

TheNLU460 may also use the device/profile state data194 to properly determine which device the input request corresponds to. For example, an incoming request such as “Alexa, stop” may be captured by one device, such as asmart watch110c(shown inFIG.12). The speech-processing system120 may receive an indication of the user profile corresponding to thesmart watch110calong with device/profile state data194 for thesmart watch110cand/or other device(s)110 associated with the particular user profile. The device/profile state data194 may indicate that a music playback device (e.g., speech-detection device110a,vehicle110e, home audio system, etc.) is capable of executing a <stopmusic> command while thesmart watch110citself may not be capable of executing such a command. TheNLU460 and/orpost-NLU ranker465 may use the device/profile state data194 to interpret the incoming request/rank hypotheses thereof to determine that the “Alexa, stop” request corresponds to a target device that is different from thesmart watch110cthat captured the input utterance. The resultingNLU results data825/885 may thus indicate that target device so that the device skill191 may send the output data to the appropriate device (e.g., speech-detection device110a,vehicle110e, home audio system, etc.) to execute the stop music command.

The NLU resultsdata825/885 may be sent (for example by theOrchestrator230 and/or another component) to thedevice skill191b. Alternatively, or in addition, theOrchestrator230 and/or another component may process theNLU results data825/885 to determine other data representing the requested device process control (e.g., <stopmusic>) where the other data is in a different form processable by thedevice skill191b. Thedevice skill191bmay then take the input data representing the user's request (where the input data may be theNLU results data825/885 or data in some other form) and may determine output data that is operable by a device component (e.g., multi-assistant component115) to cause execution of the device process control and may send that output data to theparticular device110, for example as explained above in reference toFIGS.2A and2B.

To reduce the amount of information shared between speech-processing system(s)120, thestate data194 made available to speech-processing system(s)120 may be configured to relate only to certain possible device control process(es) of a particular device. In one embodiment, the device/profile state data194 made available to speech-processing system(s)120 may only include controls executable by a device that relate to an active device process. For example, if adevice110 is outputting music, thestate data194 relative to thatdevice110 made available to a speech-processing system120 may include information related to particular controls of the music process executable by that the particular device110 (e.g., stop music, volume control, pause playback, etc.) but may not include information related to controls executable by theparticular device110, but not related to the active music playback (for example, stopping an alarm, extending a timer, etc.). In a different example, if adevice110 is outputting a beep related to an expired timer, thestate data194 relative to thatdevice110 made available to a speech-processing system120 may include information related to particular controls of the timer process executable by that the particular device110 (e.g., stop timer, extend timer, etc.) but not any information related to other, non-active, device processes. A determination of what device process(es) are active may be made by a component of therespective device110, for example by themulti-assistant component115 or other component. Themulti-assistant component115 may then select a portion of thestate data194 related to those active process(es) and send that selected portion to a speech-processing system120 (for example, as shown inFIGS.2A and2B).

In certain instances, thestate data194 may include priority data corresponding to one or more device process(es). For example, if a device has multiple controllable process(es), some of which may be active at a particular time, thestate data194 may indicate a relative priority of those processes. A speech-processing system120 may process the priority data to determine the interpretation of an input request (e.g., using NLU460) and/or the ranking of different NLU hypothes(es) of that interpretation (e.g., usingreranker890 and/or post-NLU ranker465). Example priority data may take the form of:

• • 1. Active process A
  • 2. Active process B
  • 3. Inactive process C
  • 4. Inactive process D

Thus, if adevice110 is capable of controlling four processes (A-D in the example above), two of which are active (A and B), if a user speaks a command to the device “Alexa stop” and the speech-processing system120 determines (for example using state data194) that a command of “stop” may apply to either process A or process B, an NLU hypotheses corresponding to process A may be ranked higher than an NLU hypotheses corresponding to process B as a result of the priority data. In a specific example, process A may correspond to a beeping timer and process B may correspond to music playback. Thus, if a user speaks “Alexa stop” to adevice110 without indicating what should be stopped, the system may prioritize stopping the timer and cause the timer, instead of the music playback, to be stopped.

Priority may also be impacted based on system state data195. Thus, for example, if the firstsystem state data195aof afirst system120aindicates that there is an ongoing process related to thefirst system120athat can be controlled with a “stop” command, and the user invokes thefirst system120a(for example by speaking the wakeword “Alexa” where “Alexa” invokesfirst system120a), then the process known to thefirst system120amay be prioritized. Taking the specific example above, if the process known to thefirst system120ais actually the music playback, even if the device/profile state data194 indicates a higher priority for a timer control, if thefirst system120ais performing the speech processing, its available information in the firstsystem state data195amay cause an NLU hypotheses corresponding to the music playback to be ranked higher than an NLU hypotheses corresponding to time control. Such prioritization may be configured in a variety of ways depending on the configuration ofsystem100.

In certain configurations, priority data corresponding to one or more device control processes may impact a score given to a particular NLU hypothesis by thereranker890 and/orpost-NLU ranker465. For example, priority data may be in a form of:

• • 1. Active process A [0.85]
  • 2. Active process B [0.75]
  • 3. Inactive process C [0.55]
  • 4. Inactive process D [0.45]
    The modifiers (e.g., [0.85] for process A, [0.75] for process B, etc.) may be used to adjust a score of a hypothesis corresponding to the particular process to create a resulting score to be used by thereranker890 and/orpost-NLU ranker465. In this manner the priority data may be used to determine the score and/or ranking of a hypothesis corresponding to one or more device process(es).

The device/profile state data194 may also indicate a type of a device process controllable by adevice110. For example, one type may be a dialog, another type may be a notification, another type may be a timer, another type may be media output (e.g., audio or video), another type may involve a call (e.g., video or audio communication), or the like. The type may also correspond to a component of the device used for the process such as a speaker, display, etc. The device/profile state data194 may also indicate what components of adevice110 may be active at any particular time, which may be used by a speech-processing system120 to interpret an utterance. For example, if the user says “Alexa, quiet!” the speech-processing system120 may use the device/profile state data194 to prioritize control over a device process that involves audio output.

Components of a system that may be used to perform unit selection, parametric TTS processing, and/or model-based audio synthesis are shown inFIG.9.FIG.9 is a conceptual diagram that illustrates operations for generating synthesized speech using aTTS system480, according to embodiments of the present disclosure. TheTTS system480 may receivetext data915 and process it using one ormore TTS models980 to generate synthesized speech in the form ofspectrogram data945. Avocoder990 may convert thespectrogram data945 into outputspeech audio data995, which may represent a time-domain waveform suitable for amplification and output as audio (e.g., from a loudspeaker).

TheTTS system480 may additionally receiveother input data925. Theother input data925 may include, for example, identifiers and/or labels corresponding to a desired speaker identity, voice characteristics, emotion, speech style, etc. desired for the synthesized speech. In some implementations, theother input data925 may include text tags or text metadata, that may indicate, for example, how specific words should be pronounced, for example by indicating the desired output speech quality in tags formatted according to the speech synthesis markup language (SSML) or in some other form. For example, a first text tag may be included with text marking the beginning of when text should be whispered (e.g., <begin whisper>) and a second tag may be included with text marking the end of when text should be whispered (e.g., <end whisper>). The tags may be included in thetext data915 and/or theother input data925 such as metadata accompanying a TTS request and indicating what text should be whispered (or have some other indicated audio characteristic).

TheTTS system480 may include apreprocessing component920 that can convert thetext data915 and/orother input data925 into a form suitable for processing by theTTS model980. Thetext data915 may be from, for example an application, a skill component (described further below), an NLG component, another device or source, or may be input by a user. Thetext data915 received by theTTS system480 may not necessarily be text, but may include other data (such as symbols, code, other data, etc.) that may reference text (such as an indicator of a word and/or phoneme) that is to be synthesized. Thepreprocessing component920 may transform thetext data915 into, for example, a symbolic linguistic representation, which may include linguistic context features such as phoneme data, punctuation data, syllable-level features, word-level features, and/or emotion, speaker, accent, or other features for processing by theTTS system480. The syllable-level features may include syllable emphasis, syllable speech rate, syllable inflection, or other such syllable-level features; the word-level features may include word emphasis, word speech rate, word inflection, or other such word-level features. The emotion features may include data corresponding to an emotion associated with thetext data915, such as surprise, anger, or fear. The speaker features may include data corresponding to a type of speaker, such as sex, age, or profession. The accent features may include data corresponding to an accent associated with the speaker, such as Southern, Boston, English, French, or other such accent. Style features may include a book reading style, poem reading style, a news anchor style, a sports commentator style, various singing styles, etc.

Thepreprocessing component920 may include functionality and/or components for performing text normalization, linguistic analysis, linguistic prosody generation, or other such operations. During text normalization, thepreprocessing component920 may first process thetext data915 and generate standard text, converting such things as numbers, abbreviations (such as Apt., St., etc.), symbols ($, %, etc.) into the equivalent of written out words.

During linguistic analysis, thepreprocessing component920 may analyze the language in the normalized text to generate a sequence of phonetic units corresponding to the input text. This process may be referred to as grapheme-to-phoneme conversion. Phonetic units include symbolic representations of sound units to be eventually combined and output by the system as speech. Various sound units may be used for dividing text for purposes of speech synthesis. In some implementations, theTTS model980 may process speech based on phonemes (individual sounds), half-phonemes, di-phones (the last half of one phoneme coupled with the first half of the adjacent phoneme), bi-phones (two consecutive phonemes), syllables, words, phrases, sentences, or other units. Each word may be mapped to one or more phonetic units. Such mapping may be performed using a language dictionary stored by the system, for example in a storage component. The linguistic analysis performed by thepreprocessing component920 may also identify different grammatical components such as prefixes, suffixes, phrases, punctuation, syntactic boundaries, or the like. Such grammatical components may be used by theTTS system480 to craft a natural-sounding audio waveform output. The language dictionary may also include letter-to-sound rules and other tools that may be used to pronounce previously unidentified words or letter combinations that may be encountered by theTTS system480. Generally, the more information included in the language dictionary, the higher quality the speech output.

The output of thepreprocessing component920 may be a symbolic linguistic representation, which may include a sequence of phonetic units. In some implementations, the sequence of phonetic units may be annotated with prosodic characteristics. In some implementations, prosody may be applied in part or wholly by aTTS model980. This symbolic linguistic representation may be sent to theTTS model980 for conversion into audio data (e.g., in the form of Mel-spectrograms or other frequency content data format).

TheTTS system480 may retrieve one or more previously trained and/or configuredTTS models980 from thevoice profile storage985. ATTS model980 may be, for example, a neural network architecture that may be described as interconnected artificial neurons or “cells” interconnected in layers and/or blocks. In general, neural network model architecture can be described broadly by hyperparameters that describe the number of layers and/or blocks, how many cells each layer and/or block contains, what activations functions they implement, how they interconnect, etc. A neural network model includes trainable parameters (e.g., “weights”) that indicate how much weight (e.g., in the form of an arithmetic multiplier) a cell should give to a particular input when generating an output. In some implementations, a neural network model may include other features such as a self-attention mechanism, which may determine certain parameters at run time based on inputs rather than, for example, during training based on a loss calculation. The various data that describe aparticular TTS model980 may be stored in thevoice profile storage985. ATTS model980 may represent a particular speaker identity and may be conditioned based on speaking style, emotion, etc. In some implementations, a particular speaker identity may be associated with more than oneTTS model980; for example, with a different model representing a different speaking style, language, emotion, etc. in some implementations, aparticular TTS model980 may be associated with more than one speaker identity; that is, be able to produce synthesized speech that reproduces voice characteristics of more than one character. Thus a first TTS model980amay be used to create synthesized speech for the first-speech processing system120awhile a second, different, TTS model980bmay be used to create synthesized speech for the second-speech processing system120b. In some cases, theTTS model980 may generate the desired voice characteristics based on conditioning data received or determined from thetext data915 and/or theother input data925. For example a synthesized voice of the first-speech processing system120amay be different from a synthesized voice of the second-speech processing system120b.

TheTTS system480 may, based on an indication received with thetext data915 and/orother input data925, retrieve aTTS model980 from thevoice profile storage985 and use it to process input to generate synthesized speech. TheTTS system480 may provide theTTS model980 with any relevant conditioning labels to generate synthesized speech having the desired voice characteristics. TheTTS model980 may generate spectrogram data945 (e.g., frequency content data) representing the synthesized speech, and send it to thevocoder990 for conversion into an audio signal.

TheTTS system480 may generateother output data955. Theother output data955 may include, for example, indications or instructions for handling and/or outputting the synthesized speech. For example, thetext data915 and/orother input data925 may be received along with metadata, such as SSML tags, indicating that a selected portion of thetext data915 should be louder or quieter. Thus, theother output data955 may include a volume tag that instructs thevocoder990 to increase or decrease an amplitude of the output speechaudio data995 at times corresponding to the selected portion of thetext data915. Additionally or alternatively, a volume tag may instruct a playback device to raise or lower a volume of the synthesized speech from the device's current volume level, or lower a volume of other media being output by the device (e.g., to deliver an urgent message).

Thevocoder990 may convert thespectrogram data945 generated by theTTS model980 into an audio signal (e.g., an analog or digital time-domain waveform) suitable for amplification and output as audio. Thevocoder990 may be, for example, a universal neural vocoder based on Parallel WaveNet or related model. Thevocoder990 may take as input audio data in the form of, for example, a Mel-spectrogram with80 coefficients and frequencies ranging from 50 Hz to 12 kHz. Thespeech audio data995 may be a time-domain audio format (e.g., pulse-code modulation (PCM), waveform audio format (WAV), μ-law, etc.) that may be readily converted to an analog signal for amplification and output by a loudspeaker, such as theloudspeaker112. Thespeech audio data995 may consist of, for example, 8-, 16-, or 24-bit audio having a sample rate of 16 kHz, 24 kHz, 44.1 kHz, etc. In some implementations, other bit and/or sample rates may be used.

Various machine learning techniques may be used to train and operate models to perform various steps described herein, such as user recognition, sentiment detection, image processing, dialog management, etc. Models may be trained and operated according to various machine learning techniques. Such techniques may include, for example, neural networks (such as deep neural networks and/or recurrent neural networks), inference engines, trained classifiers, etc. Examples of trained classifiers include Support Vector Machines (SVMs), neural networks, decision trees, AdaBoost (short for “Adaptive Boosting”) combined with decision trees, and random forests. Focusing on SVM as an example, SVM is a supervised learning model with associated learning algorithms that analyze data and recognize patterns in the data, and which are commonly used for classification and regression analysis. Given a set of training examples, each marked as belonging to one of two categories, an SVM training algorithm builds a model that assigns new examples into one category or the other, making it a non-probabilistic binary linear classifier. More complex SVM models may be built with the training set identifying more than two categories, with the SVM determining which category is most similar to input data. An SVM model may be mapped so that the examples of the separate categories are divided by clear gaps. New examples are then mapped into that same space and predicted to belong to a category based on which side of the gaps they fall on. Classifiers may issue a “score” indicating which category the data most closely matches. The score may provide an indication of how closely the data matches the category.

In order to apply the machine learning techniques, the machine learning processes themselves need to be trained. Training a machine learning component such as, in this case, one of the first or second models, requires establishing a “ground truth” for the training examples. In machine learning, the term “ground truth” refers to the accuracy of a training set's classification for supervised learning techniques. Various techniques may be used to train the models including backpropagation, statistical learning, supervised learning, semi-supervised learning, stochastic learning, or other known techniques.

FIG.10 is a block diagram conceptually illustrating adevice110 that may be used with the system.FIG.11 is a block diagram conceptually illustrating example components of a remote device, such as the natural languagecommand processing system120, which may assist with ASR processing, NLU processing, etc., and a skill processing component(s)125. A system (120/125) may include one or more servers. A “server” as used herein may refer to a traditional server as understood in a server/client computing structure but may also refer to a number of different computing components that may assist with the operations discussed herein. For example, a server may include one or more physical computing components (such as a rack server) that are connected to other devices/components either physically and/or over a network and is capable of performing computing operations. A server may also include one or more virtual machines that emulates a computer system and is run on one or across multiple devices. A server may also include other combinations of hardware, software, firmware, or the like to perform operations discussed herein. The server(s) may be configured to operate using one or more of a client-server model, a computer bureau model, grid computing techniques, fog computing techniques, mainframe techniques, utility computing techniques, a peer-to-peer model, sandbox techniques, or other computing techniques.

While thedevice110 may operate locally to a user (e.g., within a same environment so the device may receive inputs and playback outputs for the user) the server/system120 may be located remotely from thedevice110 as its operations may not require proximity to the user. The server/system120 may be located in an entirely different location from the device110 (for example, as part of a cloud computing system or the like) or may be located in a same environment as thedevice110 but physically separated therefrom (for example a home server or similar device that resides in a user's home or business but perhaps in a closet, basement, attic, or the like). Theprocessing system120 may also be a version of auser device110 that includes different (e.g., more) processing capabilities than other user device(s)110 in a home/office. One benefit to the server/system120 being in a user's home/business is that data used to process a command/return a response may be kept within the user's home, thus reducing potential privacy perception concerns.

Multiple systems (120/125) may be included in theoverall system100 of the present disclosure, such as one or more naturallanguage processing systems120 for performing ASR processing, one or more naturallanguage processing systems120 for performing NLU processing, one ormore skill systems125, etc. In operation, each of these systems may include computer-readable and computer-executable instructions that reside on the respective device (120/125), as will be discussed further below.

Each of these devices (110/120/125) may include one or more controllers/processors (1004/1104), which may each include a central processing unit (CPU) for processing data and computer-readable instructions, and a memory (1006/1106) for storing data and instructions of the respective device. The memories (1006/1106) may individually include volatile random access memory (RAM), non-volatile read only memory (ROM), non-volatile magnetoresistive memory (MRAM), and/or other types of memory. Each device (110/120/125) may also include a data storage component (1008/1108) for storing data and controller/processor-executable instructions. Each data storage component (1008/1108) may individually include one or more non-volatile storage types such as magnetic storage, optical storage, solid-state storage, etc. Each device (110/120/125) may also be connected to removable or external non-volatile memory and/or storage (such as a removable memory card, memory key drive, networked storage, etc.) through respective input/output device interfaces (1002/1102).

Computer instructions for operating each device (110/120/125) and its various components may be executed by the respective device's controller(s)/processor(s) (1004/1104), using the memory (1006/1106) as temporary “working” storage at runtime. A device's computer instructions may be stored in a non-transitory manner in non-volatile memory (1006/1106), storage (1008/1108), or an external device(s). Alternatively, some or all of the executable instructions may be embedded in hardware or firmware on the respective device in addition to or instead of software.

Each device (110/120/125) includes input/output device interfaces (1002/1102). A variety of components may be connected through the input/output device interfaces (1002/1102), as will be discussed further below. Additionally, each device (110/120/125) may include an address/data bus (1024/1124) for conveying data among components of the respective device. Each component within a device (110/120/125) may also be directly connected to other components in addition to (or instead of) being connected to other components across the bus (1024/1124).

Referring toFIG.10, thedevice110 may include input/output device interfaces1002 that connect to a variety of components such as an audio output component such as aspeaker112, a wired headset or a wireless headset (not illustrated), or other component capable of outputting audio. Thedevice110 may also include an audio capture component. The audio capture component may be, for example, amicrophone114 or array of microphones, a wired headset or a wireless headset (not illustrated), etc. If an array of microphones is included, approximate distance to a sound's point of origin may be determined by acoustic localization based on time and amplitude differences between sounds captured by different microphones of the array. Thedevice110 may additionally include adisplay1016 for displaying content. Thedevice110 may further include acamera1018.

Via antenna(s)1022, the input/output device interfaces1002 may connect to one ormore networks199 via a wireless local area network (WLAN) (such as Wi-Fi) radio, Bluetooth, and/or wireless network radio, such as a radio capable of communication with a wireless communication network such as a Long Term Evolution (LTE) network, WiMAX network, 3G network, 4G network, 5G network, etc. A wired connection such as Ethernet may also be supported. Through the network(s)199, the system may be distributed across a networked environment. The I/O device interface (1002/1102) may also include communication components that allow data to be exchanged between devices such as different physical servers in a collection of servers or other components.

The components of the device(s)110, the natural languagecommand processing system120, or a skill processing component(s)125 may include their own dedicated processors, memory, and/or storage. Alternatively, one or more of the components of the device(s)110, the natural languagecommand processing system120, or a skill processing component(s)125 may utilize the I/O interfaces (1002/1102), processor(s) (1004/1104), memory (1006/1106), and/or storage (1008/1108) of the device(s)110, natural languagecommand processing system120, or the skill processing component(s)125, respectively. Thus, the ASR component XXA50 may have its own I/O interface(s), processor(s), memory, and/or storage; the NLU component XXA60 may have its own I/O interface(s), processor(s), memory, and/or storage; and so forth for the various components discussed herein.

As noted above, multiple devices may be employed in a single system. In such a multi-device system, each of the devices may include different components for performing different aspects of the system's processing. The multiple devices may include overlapping components. The components of thedevice110, the natural languagecommand processing system120, and a skill processing component(s)125, as described herein, are illustrative, and may be located as a stand-alone device or may be included, in whole or in part, as a component of a larger device or system. As can be appreciated, a number of components may exist either on asystem120 and/or ondevice110. Unless expressly noted otherwise, the system version of such components may operate similarly to the device version of such components and thus the description of one version (e.g., the system version or the local version) applies to the description of the other version (e.g., the local version or system version) and vice-versa.

As illustrated inFIG.12, multiple devices (110a-110n,120,125) may contain components of the system and the devices may be connected over a network(s)199. The network(s)199 may include a local or private network or may include a wide network such as the Internet. Devices may be connected to the network(s)199 through either wired or wireless connections. For example, a speech-detection device110a, asmart phone110b, asmart watch110c, atablet computer110d, avehicle110e, a speech-detection device withdisplay110f, a display/smart television110g, a washer/dryer110h, arefrigerator110i, amicrowave110j,headphones110b/110n, etc. may be connected to the network(s)199 through a wireless service provider, over a Wi-Fi or cellular network connection, or the like. Other devices are included as network-connected support devices, such as the natural languagecommand processing system120, the skill processing component(s)125, and/or others. The support devices may connect to the network(s)199 through a wired connection or wireless connection. Networked devices may capture audio using one-or-more built-in or connected microphones or other audio capture devices, with processing performed by ASR components, NLU components, or other components of the same device or another device connected via the network(s)199, such as theASR component450, theNLU component460, etc. of the natural languagecommand processing system120.

The concepts disclosed herein may be applied within a number of different devices and computer systems, including, for example, general-purpose computing systems, speech processing systems, and distributed computing environments.

The above aspects of the present disclosure are meant to be illustrative. They were chosen to explain the principles and application of the disclosure and are not intended to be exhaustive or to limit the disclosure. Many modifications and variations of the disclosed aspects may be apparent to those of skill in the art. Persons having ordinary skill in the field of computers and speech processing should recognize that components and process steps described herein may be interchangeable with other components or steps, or combinations of components or steps, and still achieve the benefits and advantages of the present disclosure. Moreover, it should be apparent to one skilled in the art, that the disclosure may be practiced without some or all of the specific details and steps disclosed herein. Further, unless expressly stated to the contrary, features/operations/components, etc. from one embodiment discussed herein may be combined with features/operations/components, etc. from another embodiment discussed herein.

Aspects of the disclosed system may be implemented as a computer method or as an article of manufacture such as a memory device or non-transitory computer readable storage medium. The computer readable storage medium may be readable by a computer and may comprise instructions for causing a computer or other device to perform processes described in the present disclosure. The computer readable storage medium may be implemented by a volatile computer memory, non-volatile computer memory, hard drive, solid-state memory, flash drive, removable disk, and/or other media. In addition, components of system may be implemented as in firmware or hardware.

Conditional language used herein, such as, among others, “can,” “could,” “might,” “may,” “e.g.,” and the like, unless specifically stated otherwise, or otherwise understood within the context as used, is generally intended to convey that certain embodiments include, while other embodiments do not include, certain features, elements and/or steps. Thus, such conditional language is not generally intended to imply that features, elements, and/or steps are in any way required for one or more embodiments or that one or more embodiments necessarily include logic for deciding, with or without other input or prompting, whether these features, elements, and/or steps are included or are to be performed in any particular embodiment. The terms “comprising,” “including,” “having,” and the like are synonymous and are used inclusively, in an open-ended fashion, and do not exclude additional elements, features, acts, operations, and so forth. Also, the term “or” is used in its inclusive sense (and not in its exclusive sense) so that when used, for example, to connect a list of elements, the term “or” means one, some, or all of the elements in the list.

Disjunctive language such as the phrase “at least one of X, Y, Z,” unless specifically stated otherwise, is understood with the context as used in general to present that an item, term, etc., may be either X, Y, or Z, or any combination thereof (e.g., X, Y, and/or Z). Thus, such disjunctive language is not generally intended to, and should not, imply that certain embodiments require at least one of X, at least one of Y, or at least one of Z to each be present.

As used in this disclosure, the term “a” or “one” may include one or more items unless specifically stated otherwise. Further, the phrase “based on” is intended to mean “based at least in part on” unless specifically stated otherwise.

Claims (20)

What is claimed is:

1. A computer-implemented method comprising:

capturing a first utterance by a first device configured to operate with a plurality of speech-processing systems including a first speech-processing system and a second speech-processing system;

processing first audio data representing the first utterance to determine the first utterance includes a first wakeword invoking the first speech-processing system;

sending, from a first component of the first device to the first speech-processing system, the first audio data, the first component configured to operate with regard to the first speech-processing system;

receiving, by the first component from the first speech-processing system, first output data responsive to the first audio data;

sending the first output data from the first component to a second component of the first device, the second component configured to operate with regard to the plurality of speech-processing systems;

processing, by the second component, the first output data to determine a first command to terminate a process involving the first device;

processing, by the second component, state data corresponding to the first device to determine that the process was initiated in response to a command invoking the second speech-processing system;

sending, from the second component to a third component of the first device, an indication to terminate the process, the third component configured to operate with regard to the second speech-processing system; and

in response to the indication, terminating the process.

2. The computer-implemented method ofclaim 1, further comprising, prior to capturing the first utterance:

capturing a second utterance by the first device;

processing second audio data representing the second utterance to determine the second utterance includes a second wakeword invoking the second speech-processing system;

sending, from the third component to the second speech-processing system, the second audio data;

receiving, from the second speech-processing system, second output data indicating a second command to start the process;

starting the process using the first device; and

by the second component, determining the state data, wherein the state data indicates the process was initiated in response to a command invoking the second speech-processing system.

3. The computer-implemented method ofclaim 2, further comprising:

prior to receiving the first output data, sending the state data from the second component to the first speech-processing system.

4. The computer-implemented method ofclaim 1, further comprising:

determining, by the second component, that the process is active; and

after discontinuing the process, sending, from the third component to the second speech-processing system, a second indication that the process was discontinued by the first device.

5. A computer-implemented method comprising:

capturing a first utterance by a first device configured to operate with a plurality of speech-processing systems including a first speech-processing system and a second speech-processing system;

determining the first utterance corresponds to an invocation of the first speech-processing system;

sending, from the first device to the first speech-processing system, first audio data representing the first utterance;

sending, from a component of the first device to the first speech-processing system, first state data representing an active first device process of the first device, the component configured to operate with regard to the plurality of speech-processing systems;

receiving, by the component of the first device, a first command to control the first device process using the first device;

in response to the first command, executing a first operation to control the first device process; and

configuring updated state data corresponding to the first device, the updated state data reflecting execution of the first operation.

6. The computer-implemented method ofclaim 5, further comprising, prior to capturing the first utterance:

capturing a second utterance by the first device;

determining the second utterance corresponds to an invocation of the second speech-processing system;

sending, from the first device to the second speech-processing system, second audio data representing the second utterance;

receiving, by the component of the first device, a second command to initiate the first device process using the first device; and

in response to the second command, executing a second operation to initiate the first device process.

7. The computer-implemented method ofclaim 6, further comprising, after receiving the first command:

sending, from the first device to the second speech-processing system, an indication corresponding to control of the first device process.

8. The computer-implemented method ofclaim 6, further comprising:

prior to receiving the second command:

determining a first user corresponding to the second utterance,

determining the first user corresponds to a first profile, and

sending an indication of the first profile to the second speech-processing system; and

after receiving the second command and prior to receiving the first command:

determining a second user corresponding to the first utterance,

determining the second user corresponds to the first profile, and

sending an indication of the first profile to the first speech-processing system.

9. The computer-implemented method ofclaim 6, wherein:

sending the second audio data to the second speech-processing system uses a second component of the first device corresponding to the second speech-processing system; and

sending the first audio data to the first speech-processing system uses a third component of the first device corresponding to the first speech-processing system.

10. The computer-implemented method ofclaim 5, further comprising:

sending, from the first device to the first speech-processing system, second data corresponding to at least one active hardware component of the first device.

11. The computer-implemented method ofclaim 5, wherein the first state data further corresponds to a second device process controllable by the first device and the first state data indicates the first device process is active and the second device process is inactive.

12. The computer-implemented method ofclaim 11, further comprising:

determining, by the component, first priority data corresponding to the first device process;

determining, by the component, second priority data corresponding to the second device process; and

prior to receiving the first command, sending, to the first speech-processing system, the first priority data and the second priority data.

13. A device, comprising:

at least one microphone;

a first component configured to operate with regard to a plurality of speech-processing systems including a first speech-processing system and a second speech-processing system;

at least one processor; and

at least one memory comprising instructions that, when executed by the at least one processor, cause the device to:

capture a first utterance using the at least one microphone;

determine the first utterance corresponds to an invocation of the first speech-processing system;

send, to the first speech-processing system, first audio data representing the first utterance;

send, from the first component to the first speech-processing system, first state data representing an active first device process;

receive, by the first component, a first command to control the first device process;

in response to the first command, execute a first operation to control the first device process; and

configure updated state data, the updated state data reflecting execution of the first operation.

14. The device ofclaim 13, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the device to, prior to capturing the first utterance:

capture a second utterance;

determine the second utterance corresponds to an invocation of the second speech-processing system;

send, to the second speech-processing system, second audio data representing the second utterance;

receive, by the first component, a second command to initiate the first device process; and

in response to the second command, execute a second operation to initiate the first device process.

15. The device ofclaim 14, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the device to, after receiving the first command:

send, to the second speech-processing system, an indication corresponding to control of the first device process.

16. The device ofclaim 14, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the device to:

prior to receiving the second command:

determine a first user corresponding to the second utterance,

determine the first user corresponds to a first profile, and

send an indication of the first profile to the second speech-processing system; and

after receiving the second command and prior to receiving the first command:

determine a second user corresponding to the first utterance,

determine the second user corresponds to the first profile, and

send an indication of the first profile to the first speech-processing system.

17. The device ofclaim 14, wherein:

sending of the second audio data to the second speech-processing system uses a second component of the device corresponding to the second speech-processing system; and

sending of the first audio data to the first speech-processing system uses a third component of the device corresponding to the first speech-processing system.

18. The device ofclaim 13, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the device to:

send, to the first speech-processing system, second data corresponding to at least one active hardware component of the device.

19. The device ofclaim 13, wherein the first state data further corresponds to a second device process controllable by the device and the first state data indicates the first device process is active and the second device process is inactive.

20. The device ofclaim 19, wherein the at least one memory further includes instructions that, when executed by the at least one processor, further cause the device to:

determine, by the first component, first priority data corresponding to the first device process;

determine, by the first component, second priority data corresponding to the second device process; and

prior to receiving the first command, send, to the first speech-processing system, the first priority data and the second priority data.

Images